A hierarchical reserve joint clearing method and device for a power system, an electronic device, and a storage medium

Abstract

The application discloses a power system hierarchical reserve joint clearing method and device, electronic equipment and storage medium, belongs to the technical field of power system operation and dispatching, and the method comprises the steps of obtaining target period power system predicted operation data and parameters such as adjustable capacity, response time and capacity credibility of each candidate reserve resource; the renewable energy prediction penetration rate is calculated, the corresponding interval identifier is determined, and then the response, rotating and non-rotating reserve demand quantity is predicted; based on the clearing capacity decision variable, a multi-level supply and demand balance constraint, a capacity constraint and a response time constraint are constructed respectively, and a hierarchical clearing optimization model is established under the minimum operation cost objective; the target clearing capacity of each candidate reserve resource is solved, and joint clearing dispatching is carried out accordingly. Therefore, by implementing the application, the problems of insufficient fast reserve allocation causing frequency instability risk in the high penetration rate scenario and resource redundancy waste in the low penetration rate scenario existing in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of power system operation and dispatching technology, specifically to a method, apparatus, electronic device, and storage medium for the joint clearing of power system hierarchical reserves. Background Technology

[0002] With the continuous construction of new power systems and the expanding grid connection of intermittent renewable energy sources such as wind and solar power, the overall system inertia level exhibits significant dynamic changes, posing a severe challenge to the safe and stable operation of the power grid. To address the uncertainty risk of net load caused by fluctuations in renewable energy output, tiered reserve clearing has become a core component of power grid operation and dispatch. Tiered reserve clearing can coordinate the allocation of reserve capacity across multiple time scales based on the response characteristics of different reserve resources, optimizing overall dispatch costs while mitigating sudden power shortages. This is of great significance for improving the resource allocation efficiency and physical operational security of the power system.

[0003] Existing reserve clearing and scheduling methods typically rely on fixed empirical rules or static ratios for uniform clearing, failing to accurately quantify the actual impact of dynamic changes in renewable energy penetration on multi-level reserve demand. The primary reason for this problem is that existing technologies, when constructing clearing models, lack in-depth consideration of the mapping relationship between the real-time operating status of the power grid and the response speed limits of physical resources. They fail to dynamically set the maximum allowable response time constraints according to different penetration scenarios, and also neglect effective calculation of the actual physical reliability of candidate reserve resources. This deficiency of coarse demand quantification and lack of physical constraint dimension easily leads to insufficient rapid allocation of reserve resources in high-penetration, low-inertia scenarios, causing frequency instability risks, or excessive allocation of redundant resources in low-penetration scenarios, resulting in unnecessary increases in system scheduling costs. Summary of the Invention

[0004] This invention provides a method, apparatus, electronic device, and storage medium for the joint clearing of tiered reserves in power systems. It can solve the problems in the prior art where the reserve clearing method uses static ratios, which cannot accurately quantify the boundaries of multi-level reserve demand under different penetration rates, and lacks effective constraints on the physical limits of resource response time and performance reliability. This leads to insufficient rapid reserve allocation in high-penetration scenarios, causing frequency instability risks, and resource redundancy waste in low-penetration scenarios.

[0005] An embodiment of the present invention provides a method for joint clearing of tiered reserves in a power system, comprising: Acquire predicted operating data of the power system during the target period and operating characteristic parameters of multiple candidate reserve resources; operating characteristic parameters include adjustable capacity, response time parameters, and capacity reliability coefficient; Calculate the predicted renewable energy penetration rate for the target period based on the predicted operating data; determine the penetration rate interval identifier based on the predicted renewable energy penetration rate and the preset penetration rate interval division rules; input the predicted operating data and the penetration rate interval identifier into the pre-trained reserve demand prediction model, and output the second-level response reserve demand, spinning reserve demand and non-spinning reserve demand for the target period. Based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand, a multi-level supply and demand balance constraint is constructed; based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity, a capacity constraint is constructed; based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship, the upper limit of the maximum allowable response time is determined, and combined with the clearing capacity decision variables and the response time parameter, a time constraint is constructed. A hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a preset objective function for minimizing operating costs. Solve the hierarchical clearing optimization model to obtain the target clearing capacity for each candidate reserve resource among multiple candidate reserve resources; perform joint clearing scheduling on multiple candidate reserve resources according to the target clearing capacity.

[0006] Furthermore, the predicted operating data includes predicted wind power output, predicted photovoltaic power output, predicted generator output, and net incoming electricity. Calculate the predicted renewable energy penetration rate for the target period based on the predicted operational data; determine the penetration rate interval identifiers based on the predicted renewable energy penetration rate and the preset penetration rate interval division rules, including: The total predicted output of renewable energy is determined based on the predicted wind power output and the predicted photovoltaic power output. The equivalent total operating power of the system is determined based on the predicted total output power of renewable energy, the predicted output power of the generator sets, and the net received electricity. The predicted penetration rate of renewable energy is determined based on the proportion of the predicted total output power of renewable energy in the equivalent total operating power of the system. From the preset penetration rate range division rules, extract the conventional backup effectiveness boundary threshold and the rapid backup necessity boundary threshold; wherein, the conventional backup effectiveness boundary threshold is less than the rapid backup necessity boundary threshold; The predicted penetration rate of renewable energy is compared with the boundary thresholds for conventional reserve effectiveness and rapid reserve necessity, respectively. If the predicted penetration rate of renewable energy is less than the conventional reserve effectiveness boundary threshold, then the low penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the conventional reserve effectiveness boundary threshold and less than the rapid reserve necessity boundary threshold, then the medium penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the threshold of rapid backup necessity, then the high penetration rate range will be identified as the penetration rate range.

[0007] Furthermore, the reserve demand prediction model is trained in the following manner: The project acquires a training dataset for reserve demand prediction and multiple machine learning networks to be trained. The training dataset includes multiple historical time-series samples, each containing a historical multi-dimensional feature vector as input and historical reserve demand values ​​as output labels. The historical multi-dimensional feature vector includes historical operating data and historical penetration rate interval identifiers. The historical operating data includes historical wind power output, historical photovoltaic power output, historical generator output, and historical net electricity intake. The historical reserve demand values ​​include historical second-level response reserve demand, historical spinning reserve demand, and historical non-spinning reserve demand. Based on the historical penetration rate interval identifier, multiple historical time series samples are divided into multiple historical training subsets; each historical training subset corresponds to a historical penetration rate interval identifier; the multiple machine learning networks to be trained correspond one-to-one with the multiple historical training subsets. Multiple historical training subsets are input into corresponding machine learning networks to be trained for iterative training until a preset training termination condition is met, resulting in multiple trained machine learning networks. These trained machine learning networks are then aggregated into a backup demand prediction model. In each iteration, the current machine learning network to be trained performs feature mapping processing on the historical multi-dimensional feature vectors in the current historical training subset, outputting a backup demand prediction value. The prediction error between the backup demand prediction value and the corresponding historical backup demand true value is calculated using a preset loss function, generating a loss function value. A preset optimization algorithm is then used to backpropagate and update the network weight parameters of the current machine learning network to be trained based on the loss function value.

[0008] Furthermore, based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the spinning reserve demand, and the non-spinning reserve demand, a multi-level supply and demand balance constraint is constructed, including: Candidate backup resources whose response time parameter is not greater than the preset first response time threshold are classified into second-level candidate backup resources; the clearing capacity decision variables corresponding to multiple second-level candidate backup resources are summed to obtain the second-level backup supply total variable; the second-level backup supply total variable is constructed as the first-level supply and demand balance constraint condition that the second-level backup supply total variable is not less than the second-level response backup demand. Among multiple candidate reserve resources, those whose response time parameter is greater than the first response time threshold and not greater than the preset second response time threshold are identified as rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple rotating candidate reserve resources are summed to obtain the total variable of rotating reserve supply; the total variable of rotating reserve supply is not less than the rotating reserve demand to construct the second-level supply and demand balance constraint condition; wherein, the first response time threshold is less than the second response time threshold; Among multiple candidate reserve resources, those with a response time parameter greater than the second response time threshold are identified as non-rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple non-rotating candidate reserve resources are summed to obtain the total non-rotating reserve supply variable; the total non-rotating reserve supply variable is not less than the non-rotating reserve demand as a third-level supply and demand balance constraint. The first-level supply and demand balance constraints, the second-level supply and demand balance constraints, and the third-level supply and demand balance constraints are aggregated into multi-level supply and demand balance constraints.

[0009] Furthermore, based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity, capacity constraints are constructed, including: The capacity confidence coefficient of each candidate backup resource is multiplied by the corresponding adjustable capacity to obtain the corresponding effective adjustable capacity. The capacity constraint is constructed by ensuring that the clearing capacity decision variable corresponding to each candidate reserve resource is no greater than the corresponding effective adjustable capacity, and that the clearing capacity decision variable corresponding to each candidate reserve resource is no less than zero.

[0010] Furthermore, based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship, the upper limit of the maximum allowable response time is determined. Combined with the clearing capacity decision variable and the response time parameter, time constraints are constructed, including: Query the upper limit of the maximum allowable response time corresponding to the penetration rate interval identifier from the preset maximum allowable response time mapping relationship; Construct a corresponding binary decision variable for the call status of each candidate backup resource; Based on the linear scaling relationship between the preset maximum constant and the binary decision variable of the call state corresponding to each candidate backup resource, the corresponding capacity state boundary term is determined; the clearing capacity decision variable corresponding to each candidate backup resource is not greater than the corresponding capacity state boundary term, which is constructed as the first time constraint condition. The corresponding mutual exclusion state mapping term is determined based on the mutual exclusion inversion feature of the binary decision variable of the call state corresponding to each candidate backup resource; the corresponding relaxation time compensation amount is determined based on the linear scaling relationship between the maximum constant and the corresponding mutual exclusion state mapping term. The corresponding equivalent response time upper limit is determined based on the superposition and aggregation relationship between the maximum allowable response time upper limit and the corresponding relaxation time compensation amount; The response time parameter corresponding to each candidate backup resource is not greater than the corresponding equivalent response time upper limit, which is constructed as the second time constraint condition; The first and second time constraints are aggregated to form a time constraint.

[0011] Furthermore, a hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a pre-defined objective function for minimizing operating costs, including: Obtain the bid price parameters for each candidate reserve resource and the reserve shortage risk variable corresponding to each balance level in the multi-level supply and demand balance constraints; Based on the clearing capacity decision variable and the corresponding bid price parameter for each candidate reserve resource, determine the corresponding procurement cost variable; based on the aggregation relationship of the procurement cost variables corresponding to multiple candidate reserve resources, determine the total procurement cost variable of the system. Based on the mapping relationship between the preset risk penalty coefficient and the reserve shortage risk variable corresponding to each balance level, the corresponding risk penalty cost variable is determined; based on the superposition and aggregation relationship of the risk penalty cost variables corresponding to multiple balance levels, the total risk penalty cost variable of the system is determined. Based on the superposition and aggregation relationship between the total system procurement cost variable and the total system risk penalty cost variable, the total system operating cost variable is determined; with the goal of minimizing the total system operating cost variable, an objective function for minimizing operating cost is constructed. Based on multi-level supply and demand balance constraints, capacity constraints, time constraints, preset risk boundary constraints, and the objective function of minimizing operating costs, a hierarchical clearing optimization model is constructed.

[0012] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0013] One embodiment of the present invention provides a power system hierarchical reserve joint clearing device, comprising: a data acquisition module, a demand forecasting module, a model building module, and a scheduling solution module; The data acquisition module is used to acquire the predicted operating data of the power system during the target period and the operating characteristic parameters of multiple candidate reserve resources; the operating characteristic parameters include adjustable capacity, response time parameters and capacity reliability coefficient; The demand forecasting module is used to calculate the renewable energy forecast penetration rate for the target period based on the forecasted operation data; determine the penetration rate interval identifier based on the renewable energy forecast penetration rate and the preset penetration rate interval division rules; input the forecasted operation data and the penetration rate interval identifier into the pre-trained reserve demand forecasting model, and output the second-level response reserve demand, spinning reserve demand and non-spinning reserve demand for the target period. The model building module is used to construct multi-level supply and demand balance constraints based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand; construct capacity constraints based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity; determine the upper limit of the maximum allowable response time based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship, and construct time constraints in combination with the clearing capacity decision variables and response time parameters; and construct a hierarchical clearing optimization model based on the multi-level supply and demand balance constraints, capacity constraints, time constraints, and the preset objective function for minimizing operating costs. The solution scheduling module is used to solve the hierarchical clearing optimization model to obtain the target clearing capacity corresponding to each candidate reserve resource among multiple candidate reserve resources; and to perform joint clearing scheduling on multiple candidate reserve resources according to the target clearing capacity.

[0014] Based on the above method embodiments, the present invention provides corresponding electronic device embodiments.

[0015] One embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power system graded reserve joint clearing method according to any one of the above method embodiments.

[0016] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0017] One embodiment of the present invention provides a storage medium storing a computer program thereon, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described method embodiments of the power system graded reserve joint clearing method.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method, apparatus, electronic device, and storage medium for tiered reserve joint clearing in a power system. The method acquires predicted operating data of the power system during a target time period and operating characteristic parameters of multiple candidate reserve resources, including adjustable capacity, response time parameters, and capacity reliability coefficients. It calculates the predicted penetration rate of renewable energy based on the predicted operating data and determines the penetration rate interval identifier using a preset interval division rule. The predicted operating data and penetration rate interval identifiers are input into a pre-trained reserve demand prediction model, which outputs second-level response reserve demand, spinning reserve demand, and non-spinning reserve demand. Multi-level supply and demand balance constraints are constructed based on the clearing capacity decision variables of each candidate reserve resource and various reserve demand quantities. Capacity constraints are constructed using the capacity reliability coefficient and adjustable capacity, and time constraints are constructed based on the maximum allowable response time upper limit and response time parameter corresponding to the penetration rate interval. A tiered clearing optimization model is established under the objective of minimizing operating costs to solve for the target clearing capacity of each candidate reserve resource, and joint clearing scheduling is performed accordingly.

[0019] This invention solves the problem of existing technologies that use static ratios to accurately quantify the boundaries of multi-level reserve demand under different penetration rates by calculating dynamic renewable energy predicted penetration rates and using a predictive model to accurately output second-level response and spinning reserve requirements. Furthermore, this solution introduces an upper limit of the maximum allowable response time based on a penetration rate interval mapping as a time constraint into the clearing optimization model, and constructs a capacity constraint by combining it with a capacity reliability coefficient. These features compensate for the shortcomings of traditional scheduling that do not consider the physical limits of response time and the actual reliability of resource fulfillment, directly avoiding the risk of frequency instability caused by insufficient rapid reserve allocation in high-penetration scenarios and the waste of resource redundancy in low-penetration scenarios. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a power system tiered reserve joint clearing method according to an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of a power system hierarchical reserve joint clearing device provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, to address the problems in existing reserve clearing methods that use static ratios to accurately quantify multi-level reserve demand boundaries under different penetration rates, and lack effective constraints on the physical limits of resource response time and fulfillment reliability, thus leading to insufficient rapid reserve allocation in high-penetration scenarios and resulting in frequency instability risks, and resource redundancy waste in low-penetration scenarios, an embodiment of the present invention provides a graded reserve joint clearing method for power systems, comprising at least the following steps: Step S1: Obtain the predicted operating data of the power system during the target period and the operating characteristic parameters of multiple candidate reserve resources; the operating characteristic parameters include adjustable capacity, response time parameters and capacity reliability coefficient; Specifically, this involves acquiring predicted operating data of the power system during the target time period and operating characteristic parameters of multiple candidate reserve resources. As a preferred implementation, the predicted operating data includes predicted wind power output, predicted photovoltaic power output, predicted generator power output, and net incoming electricity. Before the reserve market clears, such as before the day-ahead market or real-time market clears, continuous predicted time-series data of the power network for the target time period is acquired through an energy management system or market operation platform. Predicted wind power output represents the active power expected to be injected into the power network by wind farms during the target time period. Predicted photovoltaic power output represents the active power expected to be injected into the power network by photovoltaic power plants during the target time period. Predicted generator power output represents the expected ground-state power output of conventional synchronous generators. Net incoming electricity represents the expected net energy exchange between the local power network and external networks during the target time period. Net incoming electricity is used to measure the impact of inter-regional transmission on the local power balance. Collecting rich and multi-dimensional predicted operating data lays a solid data foundation for the subsequent accurate calculation of renewable energy penetration rates.

[0024] For resource groups participating in the standby market clearing process, it is necessary to obtain the operational characteristic parameters of multiple candidate standby resources. These candidate standby resources encompass both traditional generator sets and new standby entities. New standby entities include physical entities with rapid power regulation capabilities, such as energy storage facilities or electric vehicle aggregators. Operational characteristic parameters include adjustable capacity, response time parameters, and capacity reliability coefficients. Candidate standby resources need to report multi-dimensional operational characteristic parameters to the market operator. Adjustable capacity represents the maximum standby power limit that a candidate standby resource can physically stably provide and participate in market bidding within the target time period. The response time parameter represents the time span from the moment a candidate standby resource receives a dispatch instruction to the moment its actual active power output first reaches the preset response target ratio. The preset response target ratio is configured as 90% or 100% of the target clearing capacity, used to reflect the safety baseline for the candidate standby resource's substantive energy support provision. The response time parameter is a core physical indicator characterizing the speed at which a candidate standby resource provides standby energy, and it is also the core basis for distinguishing between second-level response standby and spinning standby. The capacity reliability coefficient represents a capacity conversion parameter calculated based on the candidate standby resource's past historical performance. The capacity credibility coefficient is used to measure the true reliability of the capacity declared by candidate backup resources.

[0025] By collecting predicted operating data for the target period from multiple dimensions and refining the operating characteristic parameters of candidate reserve resources, this method overcomes the shortcomings of previous methods that relied solely on a single capacity element for extensive scheduling during reserve clearing. It achieves a forward-looking and accurate capture of the power network's operating status and a deep quantification of the physical response capabilities of reserve resources, thus providing a complete initial optimization boundary for the accurate matching and safe clearing of multi-level reserve capacity in environments with a high proportion of renewable energy access.

[0026] Step S2: Calculate the predicted renewable energy penetration rate for the target period based on the predicted operating data; determine the penetration rate interval identifier based on the predicted renewable energy penetration rate and the preset penetration rate interval division rules; input the predicted operating data and the penetration rate interval identifier into the pre-trained reserve demand prediction model, and output the second-level response reserve demand, spinning reserve demand and non-spinning reserve demand for the target period. In a preferred embodiment, the predicted operating data includes predicted wind power output, predicted photovoltaic power output, predicted generator output, and net incoming electricity. Calculate the predicted renewable energy penetration rate for the target period based on the predicted operational data; determine the penetration rate interval identifiers based on the predicted renewable energy penetration rate and the preset penetration rate interval division rules, including: The total predicted output of renewable energy is determined based on the predicted wind power output and the predicted photovoltaic power output. The equivalent total operating power of the system is determined based on the predicted total output power of renewable energy, the predicted output power of the generator sets, and the net received electricity. The predicted penetration rate of renewable energy is determined based on the proportion of the predicted total output power of renewable energy in the equivalent total operating power of the system. From the preset penetration rate range division rules, extract the conventional backup effectiveness boundary threshold and the rapid backup necessity boundary threshold; wherein, the conventional backup effectiveness boundary threshold is less than the rapid backup necessity boundary threshold; The predicted penetration rate of renewable energy is compared with the boundary thresholds for conventional reserve effectiveness and rapid reserve necessity, respectively. If the predicted penetration rate of renewable energy is less than the conventional reserve effectiveness boundary threshold, then the low penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the conventional reserve effectiveness boundary threshold and less than the rapid reserve necessity boundary threshold, then the medium penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the threshold of rapid backup necessity, then the high penetration rate range will be identified as the penetration rate range.

[0027] In a preferred embodiment, the reserve demand prediction model is trained in the following manner: The project acquires a training dataset for reserve demand prediction and multiple machine learning networks to be trained. The training dataset includes multiple historical time-series samples, each containing a historical multi-dimensional feature vector as input and historical reserve demand values ​​as output labels. The historical multi-dimensional feature vector includes historical operating data and historical penetration rate interval identifiers. The historical operating data includes historical wind power output, historical photovoltaic power output, historical generator output, and historical net electricity intake. The historical reserve demand values ​​include historical second-level response reserve demand, historical spinning reserve demand, and historical non-spinning reserve demand. Based on the historical penetration rate interval identifier, multiple historical time series samples are divided into multiple historical training subsets; each historical training subset corresponds to a historical penetration rate interval identifier; the multiple machine learning networks to be trained correspond one-to-one with the multiple historical training subsets. Multiple historical training subsets are input into corresponding machine learning networks to be trained for iterative training until a preset training termination condition is met, resulting in multiple trained machine learning networks. These trained machine learning networks are then aggregated into a backup demand prediction model. In each iteration, the current machine learning network to be trained performs feature mapping processing on the historical multi-dimensional feature vectors in the current historical training subset, outputting a backup demand prediction value. The prediction error between the backup demand prediction value and the corresponding historical backup demand true value is calculated using a preset loss function, generating a loss function value. A preset optimization algorithm is then used to backpropagate and update the network weight parameters of the current machine learning network to be trained based on the loss function value.

[0028] Specifically, before the power grid operation and scheduling enters the clearing calculation phase, the primary prerequisite is to obtain continuous predicted operational data for the target time period. In a preferred embodiment, the predicted operational data includes predicted wind power output, predicted photovoltaic power output, predicted generator power output, and net incoming electricity. Predicted wind power output represents the total active power expected to be injected into the power grid by wind power generation facilities during the target time period. Predicted photovoltaic power output represents the total active power expected to be injected into the power grid by solar power generation facilities. Predicted generator power output represents the base output power expected to be undertaken by conventional synchronous generators. Net incoming electricity reflects the energy exchange value expected when inter-regional power transmission occurs between the local power grid and external interconnected networks during the target time period.

[0029] The predicted renewable energy penetration rate for the target period is calculated based on the predicted operating data. The penetration rate interval identifier is determined according to the predicted renewable energy penetration rate and a pre-defined interval division rule. The specific process is as follows: First, the predicted output power of wind power and photovoltaic power are accumulated to determine the total predicted output power of renewable energy. Next, the total predicted output power of renewable energy, the predicted output power of generating units, and the net received electricity are normalized and integrated to determine the equivalent total operating power of the system. Then, the predicted renewable energy penetration rate is determined based on the proportion of the total predicted output power of renewable energy in the equivalent total operating power of the system. Combining the actual physical operating logic of the power network, the specific calculation formula for the predicted renewable energy penetration rate is as follows: In the formula, This indicates the projected penetration rate of renewable energy. This indicates the predicted wind power output. Indicates the predicted output power of photovoltaic power. This indicates the predicted output power of the generator set. This represents the net received electricity. By introducing the absolute value of the net received electricity to physically derive the equivalent total operating power of the system, it is possible to accurately measure the absolute impact of cross-regional power transmission on the power balance of local physical nodes.

[0030] After deriving the penetration rate values, the conventional reserve effectiveness boundary threshold and the rapid reserve necessity boundary threshold are extracted from the preset penetration rate interval division rules. The conventional reserve effectiveness boundary threshold is smaller than the rapid reserve necessity boundary threshold. Specifically, the conventional reserve effectiveness boundary threshold corresponds to the critical penetration point when the maximum frequency change rate curve of the power network shows a significant non-linear growth trend, representing the extreme physical state where the rotational inertia retained by conventional synchronous generators is still sufficient to maintain global frequency stability. The rapid reserve necessity boundary threshold corresponds to the critical penetration point when the maximum frequency change rate of the power network first crosses the safe operation red line, representing the critical physical state boundary where the power network inertia is extremely low and second-level active response resources must be introduced to prevent a deep frequency drop. After extracting the conventional reserve effectiveness boundary threshold and the rapid reserve necessity boundary threshold, the predicted renewable energy penetration rate is compared with both the conventional reserve effectiveness boundary threshold and the rapid reserve necessity boundary threshold.

[0031] If the predicted penetration rate of renewable energy is less than the conventional reserve effectiveness threshold, it indicates that the power grid has sufficient rotational inertia reserves, and the reserve resources provided by conventional power generation equipment are fully sufficient to cope with daily frequency fluctuations. In this case, the low penetration rate range is identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the conventional reserve effectiveness threshold but less than the rapid reserve necessity threshold, it indicates that the power grid is in a transitional operation phase with a gradual decrease in inertia level, and the demand for non-rotating reserve resources is showing an upward trend. In this case, the medium penetration rate range is identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the rapid reserve necessity threshold, it indicates that a significant reduction in synchronous power generation equipment has led to a sharp increase in the risk of frequency instability, and the power grid is in a high-risk operating state, requiring mandatory inclusion of ultra-fast response capabilities. In this case, the high penetration rate range is identified as the penetration rate range.

[0032] In a preferred embodiment, a reserve demand prediction model is trained through a complete closed-loop machine learning process. A reserve demand prediction training dataset and multiple machine learning networks to be trained are obtained. These machine learning networks can be constructed using an extreme gradient boosting tree architecture, leveraging the node splitting mechanism of the tree model to effectively capture the complex nonlinear mapping relationship between high-dimensional input features and multi-level reserve demand. The reserve demand prediction training dataset includes multiple historical time-series samples. These samples contain historical multi-dimensional feature vectors as input and historical reserve demand real values ​​as output labels. The historical multi-dimensional feature vectors include historical wind power output, historical photovoltaic power output, historical generator output, historical net electricity intake, and historical penetration rate interval identifiers. The historical reserve demand real values ​​include historical second-level response reserve demand, historical spinning reserve demand, and historical non-spinning reserve demand. These historical reserve demand real values ​​are derived from dynamic retrospective analysis of historical real power disturbance events. The historical second-level response reserve demand corresponds to the total amount of active power actually added to the power grid within the first second after a physical disturbance event occurs. Historical non-spinning reserve demand corresponds to the difference between the total active power increase in the power grid within 30 seconds after a physical disturbance event and the historical second-level response reserve demand. Historical spinning reserve demand corresponds to the difference between the total active power increase in the power grid within 5 minutes after a physical disturbance event and the historical non-spinning reserve demand and the historical second-level response reserve demand.

[0033] After obtaining structured sample data, multiple historical time-series samples are divided into multiple historical training subsets based on historical penetration rate interval identifiers. Each historical training subset corresponds to a historical penetration rate interval identifier. Multiple machine learning networks to be trained correspond one-to-one with multiple historical training subsets. Introducing prior physical interval segmentation logic enables independently allocated machine learning network branches to focus on capturing the nonlinear feature evolution patterns under specific physical operating environments. Multiple historical training subsets are input into the corresponding machine learning networks to be trained for iterative training until a preset training termination condition is met, resulting in multiple trained machine learning networks. These trained machine learning networks are then aggregated into a backup demand prediction model. In each micro-iterative calculation cycle, the current machine learning network to be trained performs feature mapping processing on the historical multi-dimensional feature vectors in the current historical training subset, calculates and outputs backup demand prediction values. Subsequently, a preset loss function rigorously evaluates the prediction error between the backup demand prediction value and the corresponding historical backup demand true value, thereby generating an intuitive loss function value. Using a preset optimization algorithm, the network weight parameters of the current machine learning network to be trained are updated via backpropagation along the direction of reducing the error gradient, based on the loss function value. The preset training termination conditions are configured as follows: the number of iterations of network weight parameters reaches a preset upper limit, or the loss function value drops to a preset convergence threshold range; the preset loss function is configured as the mean squared error loss function; and the preset optimization algorithm is configured as the adaptive moment estimation optimization algorithm.

[0034] By implementing data feature calculations and multi-dimensional segmented prediction steps, we have achieved accurate and forward-looking quantification of the differentiated reserve demand boundaries under various renewable energy access scenarios. This has completely resolved the technical bottleneck that traditional scheduling strategies cannot accurately distinguish physical response speed requirements. It not only effectively avoids the risk of grid collapse caused by insufficient procurement of rapid support resources in high-proportion renewable energy access environments, but also significantly reduces the reserve redundancy procurement costs in low-proportion renewable energy access environments.

[0035] Step S3: Based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand, construct multi-level supply and demand balance constraints; based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity, construct capacity constraints; determine the upper limit of the maximum allowable response time based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship, and construct time constraints by combining the clearing capacity decision variables and response time parameters; construct a hierarchical clearing optimization model based on the multi-level supply and demand balance constraints, capacity constraints, time constraints, and the preset operating cost minimization objective function. In a preferred embodiment, a multi-level supply and demand balance constraint is constructed based on the clearing capacity decision variables of the plurality of candidate reserve resources, the second-level response reserve demand, the spinning reserve demand, and the non-spinning reserve demand, including: Candidate backup resources whose response time parameter is not greater than the preset first response time threshold are classified into second-level candidate backup resources; the clearing capacity decision variables corresponding to multiple second-level candidate backup resources are summed to obtain the second-level backup supply total variable; the second-level backup supply total variable is constructed as the first-level supply and demand balance constraint condition that the second-level backup supply total variable is not less than the second-level response backup demand. Among multiple candidate reserve resources, those whose response time parameter is greater than the first response time threshold and not greater than the preset second response time threshold are identified as rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple rotating candidate reserve resources are summed to obtain the total variable of rotating reserve supply; the total variable of rotating reserve supply is not less than the rotating reserve demand to construct the second-level supply and demand balance constraint condition; wherein, the first response time threshold is less than the second response time threshold; Among multiple candidate reserve resources, those with a response time parameter greater than the second response time threshold are identified as non-rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple non-rotating candidate reserve resources are summed to obtain the total non-rotating reserve supply variable; the total non-rotating reserve supply variable is not less than the non-rotating reserve demand as a third-level supply and demand balance constraint. The first-level supply and demand balance constraints, the second-level supply and demand balance constraints, and the third-level supply and demand balance constraints are aggregated into multi-level supply and demand balance constraints.

[0036] In a preferred embodiment, capacity constraints are constructed based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity, including: The capacity confidence coefficient of each candidate backup resource is multiplied by the corresponding adjustable capacity to obtain the corresponding effective adjustable capacity. The capacity constraint is constructed by ensuring that the clearing capacity decision variable corresponding to each candidate reserve resource is no greater than the corresponding effective adjustable capacity, and that the clearing capacity decision variable corresponding to each candidate reserve resource is no less than zero.

[0037] In a preferred embodiment, the upper limit of the maximum allowable response time is determined based on the penetration rate interval identifier and a preset maximum allowable response time mapping relationship. Combined with the clearing capacity decision variable and the response time parameter, time constraints are constructed, including: Query the upper limit of the maximum allowable response time corresponding to the penetration rate interval identifier from the preset maximum allowable response time mapping relationship; Construct a corresponding binary decision variable for the call status of each candidate backup resource; Based on the linear scaling relationship between the preset maximum constant and the binary decision variable of the call state corresponding to each candidate backup resource, the corresponding capacity state boundary term is determined; the clearing capacity decision variable corresponding to each candidate backup resource is not greater than the corresponding capacity state boundary term, which is constructed as the first time constraint condition. The corresponding mutual exclusion state mapping term is determined based on the mutual exclusion inversion feature of the binary decision variable of the call state corresponding to each candidate backup resource; the corresponding relaxation time compensation amount is determined based on the linear scaling relationship between the maximum constant and the corresponding mutual exclusion state mapping term. The corresponding equivalent response time upper limit is determined based on the superposition and aggregation relationship between the maximum allowable response time upper limit and the corresponding relaxation time compensation amount; The response time parameter corresponding to each candidate backup resource is not greater than the corresponding equivalent response time upper limit, which is constructed as the second time constraint condition; The first and second time constraints are aggregated to form a time constraint.

[0038] In a preferred embodiment, a hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a preset objective function for minimizing operating costs, including: Obtain the bid price parameters for each candidate reserve resource and the reserve shortage risk variable corresponding to each balance level in the multi-level supply and demand balance constraints; Based on the clearing capacity decision variable and the corresponding bid price parameter for each candidate reserve resource, determine the corresponding procurement cost variable; based on the aggregation relationship of the procurement cost variables corresponding to multiple candidate reserve resources, determine the total procurement cost variable of the system. Based on the mapping relationship between the preset risk penalty coefficient and the reserve shortage risk variable corresponding to each balance level, the corresponding risk penalty cost variable is determined; based on the superposition and aggregation relationship of the risk penalty cost variables corresponding to multiple balance levels, the total risk penalty cost variable of the system is determined. Based on the superposition and aggregation relationship between the total system procurement cost variable and the total system risk penalty cost variable, the total system operating cost variable is determined; with the goal of minimizing the total system operating cost variable, an objective function for minimizing operating cost is constructed. Based on multi-level supply and demand balance constraints, capacity constraints, time constraints, preset risk boundary constraints, and the objective function of minimizing operating costs, a hierarchical clearing optimization model is constructed.

[0039] Specifically, after predicting various types of reserve demands, a rigorous mathematical programming framework is needed to find the optimal resource allocation scheme. A continuous clearing capacity decision variable is introduced to represent the actual winning reserve capacity allocated to each candidate reserve resource. In a preferred embodiment, multi-level supply and demand balance constraints are constructed based on the clearing capacity decision variables of multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand. The specific process involves hierarchical comparison and aggregation of response time parameters. Candidate reserve resources whose response time parameters are not greater than a preset first response time threshold are classified as second-level candidate reserve resources. The first response time threshold represents the physical time boundary of ultra-fast response resources. Subsequently, the clearing capacity decision variables corresponding to multiple second-level candidate reserve resources are summed to obtain the second-level reserve supply total variable. The second-level reserve supply total variable being no less than the second-level response reserve demand constitutes the first-level supply and demand balance constraint.

[0040] Along the dimension of increasing response time, candidate reserve resources whose response time parameters are greater than a first response time threshold but not greater than a preset second response time threshold are identified as rotating candidate reserve resources. The second response time threshold represents the physical time boundary of minute-level conventional response resources, and the first response time threshold is strictly less than the second response time threshold. The first response time threshold is set based on the time point of the minimum frequency drop inertia after the power network is actually subjected to power disturbance; the second response time threshold is set based on the physical handover time scale of the primary and secondary frequency regulation control actions of the power network. The clearing capacity decision variables corresponding to multiple rotating candidate reserve resources are summed to obtain the total rotating reserve supply variable. The requirement that the total rotating reserve supply variable is not less than the rotating reserve demand is constructed as the second-level supply-demand balance constraint. Similarly, candidate reserve resources whose response time parameters are greater than the second response time threshold are identified as non-rotating candidate reserve resources. The clearing capacity decision variables corresponding to multiple non-rotating candidate reserve resources are summed to obtain the total non-rotating reserve supply variable. The requirement that the total non-rotating reserve supply variable is not less than the non-rotating reserve demand is constructed as the third-level supply-demand balance constraint. Finally, the first-level supply and demand balance constraints, the second-level supply and demand balance constraints, and the third-level supply and demand balance constraints are aggregated into multi-level supply and demand balance constraints to ensure that various types of reserve needs are accurately met.

[0041] In a preferred embodiment, capacity constraints are constructed based on clearing capacity decision variables, capacity reliability coefficients, and adjustable capacity. The effective adjustable capacity is obtained by multiplying the capacity reliability coefficient of each candidate reserve resource by its corresponding adjustable capacity. Capacity constraints are constructed where the clearing capacity decision variable for each candidate reserve resource is no greater than the corresponding effective adjustable capacity, and the clearing capacity decision variable for each candidate reserve resource is no less than zero. Introducing a capacity reliability coefficient to discount the physical declared capacity effectively avoids excessive reliance on low-reliability resources with poor historical performance in security scheduling, forcing optimization calculations to automatically distribute procurement plans to physical entities with high performance value.

[0042] In a preferred embodiment, the upper limit of the maximum allowable response time is determined based on the penetration rate interval identifier and a preset maximum allowable response time mapping relationship. Time constraints are then constructed by combining the clearing capacity decision variable and the response time parameter. The upper limit of the maximum allowable response time corresponding to the penetration rate interval identifier is queried from the preset maximum allowable response time mapping relationship. The preset maximum allowable response time mapping relationship is configured such that the higher the renewable energy penetration level represented by the penetration rate interval identifier, the shorter the corresponding maximum allowable response time upper limit. This forces high-proportion renewable energy operating conditions to prioritize matching candidate backup resources with faster response speeds at the physical constraint level. A corresponding binary decision variable for the call status is constructed for each candidate backup resource. The binary decision variable for the call status takes a value of zero or one, used to characterize whether the resource is substantially called. Based on the linear scaling relationship between the preset maximum constant and the binary decision variable for the call status corresponding to each candidate backup resource, the corresponding capacity status boundary term is determined. The clearing capacity decision variable corresponding to each candidate backup resource is no greater than the corresponding capacity status boundary term, which is constructed as the first time constraint condition. The first time constraint condition forcibly locks the physical linkage logic between capacity allocation and call status. The preset maximum constant is configured as a positive real number that is strictly greater than the sum of the effective adjustable capacity of all candidate backup resources within the power system and the maximum value of the corresponding response time parameter, so as to ensure that when the binary decision variable of the call state takes a zero value or a one value, the corresponding linear constraint condition can automatically achieve mathematical relaxation failure.

[0043] Based on the mutual exclusion inversion characteristics of the binary decision variables for the call state corresponding to each candidate backup resource, the corresponding mutual exclusion state mapping term is determined. Based on the linear scaling relationship between the maximum constant and the corresponding mutual exclusion state mapping term, the corresponding relaxation time compensation amount is determined. Based on the superposition and aggregation relationship between the maximum allowable response time upper limit and the corresponding relaxation time compensation amount, the corresponding equivalent response time upper limit is determined. The second time constraint condition is constructed by ensuring that the response time parameter corresponding to each candidate backup resource is no greater than the corresponding equivalent response time upper limit. When the resource is not allocated capacity, the relaxation time compensation amount reaches its maximum value, thus automatically invalidating the constraint; when the resource is allocated capacity, the relaxation time compensation amount returns to zero, forcing the response time parameter to be precisely within the maximum allowable response time upper limit. Finally, the first and second time constraints are aggregated to construct the final time constraint condition.

[0044] In a preferred embodiment, a hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a preset objective function for minimizing operating costs. The bid price parameters for each candidate reserve resource and the reserve shortage risk variable corresponding to each balance level in the multi-level supply and demand balance constraints are obtained. The reserve shortage risk variable is used to quantify the positive exposure of the difference between various reserve demands and total adoption. Based on the clearing capacity decision variable and the corresponding bid price parameters for each candidate reserve resource, the corresponding procurement cost variable is determined. Based on the aggregation relationship of the procurement cost variables corresponding to multiple candidate reserve resources, the total system procurement cost variable is determined.

[0045] The variable representing the risk of reserve shortage is constrained by preset risk boundary conditions. Since the nonlinear risk logic cannot be directly input into the mixed-integer programming solution framework, the risk boundary constraints are physically equivalently constructed through a series of linear inequalities. The specific calculation formula is as follows: In the formula, This represents a variable indicating the risk of a shortage of reserves. This represents the total reserve requirement at the corresponding balancing level; This represents the total reserve supply variable corresponding to the balancing level; This represents the preset maximum constant; Let represent the binary decision variable for a state of ample capacity. A value of zero indicates sufficient supply, and a value of one indicates insufficient supply. Using the above system of inequalities, a positive difference is locked in when supply is insufficient, and the risk is forced to zero when supply is sufficient.

[0046] Based on the mapping relationship between the preset risk penalty coefficient and the reserve shortage risk variable corresponding to each balance level, the corresponding risk penalty cost variable is determined. The preset risk penalty coefficient is quantified based on the power network load shedding value or the preset administrative scarcity electricity price in the electricity market to truly reflect the extreme economic loss caused by reserve capacity shortage to the stable operation of the power network. Based on the superposition and aggregation relationship of the risk penalty cost variables corresponding to multiple balance levels, the total system risk penalty cost variable is determined. Based on the superposition and aggregation relationship between the total system procurement cost variable and the total system risk penalty cost variable, the total system operating cost variable is determined. With the goal of minimizing the total system operating cost variable, an objective function for minimizing operating costs is constructed. Finally, based on multi-level supply and demand balance constraints, capacity constraints, time constraints, preset risk boundary constraints, and the objective function for minimizing operating costs, a complete hierarchical clearing optimization model is constructed.

[0047] By implementing multi-dimensional constraint construction and risk penalty linearization derivation, the performance constraint and risk cost measurement mechanism are creatively integrated into the objective function. This enables the optimization model to accurately screen and only allow high-quality resources that perfectly match the physical characteristics of the current penetration rate to win the bid, greatly improving the stability of the global defense and the scheduling economy under extreme operating conditions.

[0048] Step S4: Solve the hierarchical clearing optimization model to obtain the target clearing capacity for each candidate reserve resource among multiple candidate reserve resources; perform joint clearing scheduling on multiple candidate reserve resources according to the target clearing capacity.

[0049] Specifically, after constructing the hierarchical clearing optimization model, a mixed-integer linear programming algorithm is introduced to jointly optimize the multivariate variables within the model. During the calculation, based on branch-and-bound logic or interior-point optimization logic, node exploration and spatial pruning are performed on integer variables, including binary decision variables for capacity sufficiency and call states. Simultaneously, gradient optimization is performed on the clearing capacity decision variables with continuous properties. After continuous iterative solving, a numerical solution is output that satisfies all multi-level supply-demand balance constraints, capacity constraints, and time constraints, and minimizes the total system operating cost variable. The optimal allocation value for each candidate reserve resource is extracted from the numerical solution of the multivariate variables, and this optimal allocation value is determined as the target clearing capacity.

[0050] After obtaining the target clearing capacity, physical control commands are issued to multiple candidate reserve resources based on the target clearing capacity to execute joint clearing scheduling. The scheduling control center sends automatic generation control signals carrying the target clearing capacity value to the corresponding physical resource nodes via the communication network. Upon receiving the scheduling signal, the physical generation entity or energy storage entity adjusts the active power output reference value of the underlying inverter or generator speed governor according to the signal requirements, thereby providing precise reserve power support within the set physical time boundaries.

[0051] After executing joint clearing scheduling and completing the physical operation of the target time period, dynamic settlement amount adjustments are required based on the actual physical performance of candidate reserve resources. The actual response time of the candidate reserve resources during the provision of reserve power is obtained. Actual response time represents the actual physical time elapsed from when the candidate reserve resource receives the scheduling instruction until its active power output reaches the target clearing capacity set ratio. The actual response time is compared with the previously obtained response time parameter. Under operating conditions where the actual response time is greater than the response time parameter, the response latency rate is calculated. The specific formula for calculating the response latency rate is as follows: In the formula, Indicates response latency rate. Indicates the actual response time. This represents the response time parameter.

[0052] After deriving the response latency rate, and combining the previously obtained bid price parameters and target clearing capacity, the performance penalty amount for default on candidate standby resource obligations is determined. The performance penalty amount is used to reduce the economic benefits of standby entities that have not strictly adhered to their physical response time commitments. The specific calculation formula for the performance penalty amount is as follows: In the formula, This indicates the amount of the performance penalty. Indicates the target clearing capacity. Indicates the parameters of the declared price. This represents the preset performance penalty weight. The performance penalty weight is preset by the electricity market operator based on the power grid safety operation assessment standards. After calculating the performance penalty amount, the actual settlement amount for candidate reserve resources is determined based on the target clearing capacity, the declared price parameters, and the performance penalty amount. The specific formula for calculating the actual settlement amount is as follows: In the formula expression, This represents the actual total settlement amount. Under operating conditions where the actual response time does not exceed the response time parameter, the performance penalty amount is directly set to zero, and the actual total settlement amount of the corresponding candidate standby resources is equal to the product of the target clearing capacity and the declared price parameter.

[0053] By implementing precise solutions from operations research models and issuing physical dispatch instructions, and superimposing a dynamic performance penalty settlement mechanism based on actual response performance, the optimal economic allocation of multi-level reserve capacity in complex power networks is achieved. By leveraging a strict economic settlement closed loop, the supply side of reserve resources is forced to improve the reliability of physical performance, fundamentally strengthening the operational safety defense line of the power network in the face of drastic fluctuations in net load.

[0054] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0055] like Figure 2 As shown, an embodiment of the present invention provides a power system hierarchical reserve joint clearing device, including: a data acquisition module, a demand forecasting module, a model building module, and a scheduling solution module; The data acquisition module is used to acquire the predicted operating data of the power system during the target period and the operating characteristic parameters of multiple candidate reserve resources; the operating characteristic parameters include adjustable capacity, response time parameters and capacity reliability coefficient; The demand forecasting module is used to calculate the renewable energy forecast penetration rate for the target period based on the forecasted operation data; determine the penetration rate interval identifier based on the renewable energy forecast penetration rate and the preset penetration rate interval division rules; input the forecasted operation data and the penetration rate interval identifier into the pre-trained reserve demand forecasting model, and output the second-level response reserve demand, spinning reserve demand and non-spinning reserve demand for the target period. The model building module is used to construct multi-level supply and demand balance constraints based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand; construct capacity constraints based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity; determine the upper limit of the maximum allowable response time based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship, and construct time constraints in combination with the clearing capacity decision variables and response time parameters; and construct a hierarchical clearing optimization model based on the multi-level supply and demand balance constraints, capacity constraints, time constraints, and the preset objective function for minimizing operating costs. The solution scheduling module is used to solve the hierarchical clearing optimization model to obtain the target clearing capacity corresponding to each candidate reserve resource among multiple candidate reserve resources; and to perform joint clearing scheduling on multiple candidate reserve resources according to the target clearing capacity.

[0056] It should be noted that the embodiments of the apparatus described above correspond to the embodiments of the present invention described above, and can realize the power system graded reserve joint clearing method described in any one of the above embodiments of the present invention. Furthermore, the embodiments of the apparatus described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the apparatus embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without creative effort.

[0057] Based on the above-described method embodiments of the present invention, a corresponding embodiment of an electronic device is provided.

[0058] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power system graded reserve joint clearing method according to any one of the present invention, or, when the processor executes the computer program, it implements the functions of each module in the above-described device embodiments.

[0059] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0060] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0061] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0062] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0063] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments; Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is executed, the device where the storage medium is located controls the execution of any of the above-described power system graded reserve joint clearing methods of the present invention.

[0064] The aforementioned storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0065] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0066] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for joint clearing of tiered reserves in a power system, characterized in that, include: Acquire predicted operating data of the power system during the target period and operating characteristic parameters of multiple candidate reserve resources; operating characteristic parameters include adjustable capacity, response time parameters, and capacity reliability coefficient; Calculate the predicted penetration rate of renewable energy for the target period based on the predicted operational data; determine the penetration rate interval identifier based on the predicted penetration rate of renewable energy and the preset penetration rate interval division rules; Input the predicted operational data and penetration rate interval identifier into the pre-trained reserve demand prediction model, and output the second-level response reserve demand, rotating reserve demand and non-rotating reserve demand for the target period. Based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand, a multi-level supply and demand balance constraint is constructed. Based on the clearing capacity decision variables, the capacity credibility coefficient, and the adjustable capacity, capacity constraints are constructed. The upper limit of the maximum allowable response time is determined based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship. Time constraints are then constructed by combining the clearing capacity decision variables and response time parameters. A hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a preset objective function for minimizing operating costs. Solve the hierarchical clearing optimization model to obtain the target clearing capacity for each candidate reserve resource among multiple candidate reserve resources; Based on the target clearing capacity, a joint clearing schedule is performed on multiple candidate reserve resources.

2. The power system tiered reserve joint clearing method as described in claim 1, characterized in that, The predicted operating data includes predicted wind power output, predicted photovoltaic power output, predicted generator output, and net incoming electricity. Calculate the predicted penetration rate of renewable energy for the target period based on the predicted operational data; The penetration rate interval identifiers are determined based on the predicted penetration rate of renewable energy and the pre-defined penetration rate interval division rules, including: The total predicted output of renewable energy is determined based on the predicted wind power output and the predicted photovoltaic power output. The equivalent total operating power of the system is determined based on the predicted total output power of renewable energy, the predicted output power of the generator sets, and the net received electricity. The predicted penetration rate of renewable energy is determined based on the proportion of the predicted total output power of renewable energy in the equivalent total operating power of the system. From the preset penetration rate range division rules, extract the conventional backup effectiveness boundary threshold and the rapid backup necessity boundary threshold; wherein, the conventional backup effectiveness boundary threshold is less than the rapid backup necessity boundary threshold; The predicted penetration rate of renewable energy is compared with the boundary thresholds for conventional reserve effectiveness and rapid reserve necessity, respectively. If the predicted penetration rate of renewable energy is less than the conventional reserve effectiveness boundary threshold, then the low penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the conventional reserve effectiveness boundary threshold and less than the rapid reserve necessity boundary threshold, then the medium penetration rate range will be identified as the penetration rate range. If the predicted penetration rate of renewable energy is greater than or equal to the threshold of rapid backup necessity, then the high penetration rate range will be identified as the penetration rate range.

3. The power system tiered reserve joint clearing method as described in claim 2, characterized in that, The reserve demand prediction model is trained using the following method: The project acquires a training dataset for reserve demand prediction and multiple machine learning networks to be trained. The training dataset includes multiple historical time-series samples, each containing a historical multi-dimensional feature vector as input and historical reserve demand values ​​as output labels. The historical multi-dimensional feature vector includes historical operating data and historical penetration rate interval identifiers. The historical operating data includes historical wind power output, historical photovoltaic power output, historical generator output, and historical net electricity intake. The historical reserve demand values ​​include historical second-level response reserve demand, historical spinning reserve demand, and historical non-spinning reserve demand. Based on the historical penetration rate interval identifier, multiple historical time series samples are divided into multiple historical training subsets; each historical training subset corresponds to a historical penetration rate interval identifier; the multiple machine learning networks to be trained correspond one-to-one with the multiple historical training subsets. Multiple historical training subsets are input into corresponding machine learning networks to be trained for iterative training until a preset training termination condition is met, resulting in multiple trained machine learning networks. These trained machine learning networks are then aggregated into a backup demand prediction model. In each iteration, the current machine learning network to be trained performs feature mapping processing on the historical multi-dimensional feature vectors in the current historical training subset, outputting a backup demand prediction value. The prediction error between the backup demand prediction value and the corresponding historical backup demand true value is calculated using a preset loss function, generating a loss function value. A preset optimization algorithm is then used to backpropagate and update the network weight parameters of the current machine learning network to be trained based on the loss function value.

4. The power system tiered reserve joint clearing method as described in claim 3, characterized in that, Based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the spinning reserve demand, and the non-spinning reserve demand, a multi-level supply and demand balance constraint is constructed, including: Candidate backup resources whose response time parameter is not greater than the preset first response time threshold are classified into second-level candidate backup resources; the clearing capacity decision variables corresponding to multiple second-level candidate backup resources are summed to obtain the second-level backup supply total variable; the second-level backup supply total variable is constructed as the first-level supply and demand balance constraint condition that the second-level backup supply total variable is not less than the second-level response backup demand. Among multiple candidate reserve resources, those whose response time parameter is greater than the first response time threshold and not greater than the preset second response time threshold are identified as rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple rotating candidate reserve resources are summed to obtain the total variable of rotating reserve supply; the total variable of rotating reserve supply is not less than the rotating reserve demand to construct the second-level supply and demand balance constraint condition; wherein, the first response time threshold is less than the second response time threshold; Among multiple candidate reserve resources, those with a response time parameter greater than the second response time threshold are identified as non-rotating candidate reserve resources; the clearing capacity decision variables corresponding to multiple non-rotating candidate reserve resources are summed to obtain the total non-rotating reserve supply variable; the total non-rotating reserve supply variable is not less than the non-rotating reserve demand as a third-level supply and demand balance constraint. The first-level supply and demand balance constraints, the second-level supply and demand balance constraints, and the third-level supply and demand balance constraints are aggregated into multi-level supply and demand balance constraints.

5. The power system tiered reserve joint clearing method as described in claim 4, characterized in that, Based on the clearing capacity decision variables, the capacity reliability coefficient, and the adjustable capacity, capacity constraints are constructed, including: The capacity confidence coefficient of each candidate backup resource is multiplied by the corresponding adjustable capacity to obtain the corresponding effective adjustable capacity. The capacity constraint is constructed by ensuring that the clearing capacity decision variable corresponding to each candidate reserve resource is no greater than the corresponding effective adjustable capacity, and that the clearing capacity decision variable corresponding to each candidate reserve resource is no less than zero.

6. The power system tiered reserve joint clearing method as described in claim 5, characterized in that, Based on the penetration rate range identifier and the preset maximum allowable response time mapping relationship, the upper limit of the maximum allowable response time is determined. Combined with the clearing capacity decision variable and the response time parameter, time constraints are constructed, including: Query the upper limit of the maximum allowable response time corresponding to the penetration rate interval identifier from the preset maximum allowable response time mapping relationship; Construct a corresponding binary decision variable for the call status of each candidate backup resource; Based on the linear scaling relationship between the preset maximum constant and the binary decision variable of the call state corresponding to each candidate backup resource, the corresponding capacity state boundary term is determined; the clearing capacity decision variable corresponding to each candidate backup resource is not greater than the corresponding capacity state boundary term, which is constructed as the first time constraint condition. The corresponding mutual exclusion state mapping term is determined based on the mutual exclusion inversion feature of the binary decision variable of the call state corresponding to each candidate backup resource; the corresponding relaxation time compensation amount is determined based on the linear scaling relationship between the maximum constant and the corresponding mutual exclusion state mapping term. The corresponding equivalent response time upper limit is determined based on the superposition and aggregation relationship between the maximum allowable response time upper limit and the corresponding relaxation time compensation amount; The response time parameter corresponding to each candidate backup resource is not greater than the corresponding equivalent response time upper limit, which is constructed as the second time constraint condition; The first and second time constraints are aggregated to form a time constraint.

7. The power system tiered reserve joint clearing method as described in claim 6, characterized in that, A hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a pre-defined objective function for minimizing operating costs, including: Obtain the bid price parameters for each candidate reserve resource and the reserve shortage risk variable corresponding to each balance level in the multi-level supply and demand balance constraints; Based on the clearing capacity decision variable and the corresponding bid price parameter for each candidate reserve resource, determine the corresponding procurement cost variable; based on the aggregation relationship of the procurement cost variables corresponding to multiple candidate reserve resources, determine the total procurement cost variable of the system. Based on the mapping relationship between the preset risk penalty coefficient and the reserve shortage risk variable corresponding to each balance level, the corresponding risk penalty cost variable is determined; based on the superposition and aggregation relationship of the risk penalty cost variables corresponding to multiple balance levels, the total risk penalty cost variable of the system is determined. Based on the superposition and aggregation relationship between the total system procurement cost variable and the total system risk penalty cost variable, the total system operating cost variable is determined; with the goal of minimizing the total system operating cost variable, an objective function for minimizing operating cost is constructed. Based on multi-level supply and demand balance constraints, capacity constraints, time constraints, preset risk boundary constraints, and the objective function of minimizing operating costs, a hierarchical clearing optimization model is constructed.

8. A power system tiered reserve joint clearing device, characterized in that, include: The module comprises a data acquisition module, a demand forecasting module, a model building module, and a solution scheduling module. The data acquisition module is used to acquire the predicted operating data of the power system during the target period and the operating characteristic parameters of multiple candidate reserve resources; the operating characteristic parameters include adjustable capacity, response time parameters and capacity reliability coefficient; The demand forecasting module is used to calculate the predicted penetration rate of renewable energy for the target period based on the predicted operating data; and to determine the penetration rate interval identifier based on the predicted penetration rate of renewable energy and the preset penetration rate interval division rules. Input the predicted operational data and penetration rate interval identifier into the pre-trained reserve demand prediction model, and output the second-level response reserve demand, rotating reserve demand and non-rotating reserve demand for the target period. The model building module is used to construct multi-level supply and demand balance constraints based on the clearing capacity decision variables of the multiple candidate reserve resources, the second-level response reserve demand, the rotating reserve demand, and the non-rotating reserve demand. Based on the clearing capacity decision variables, the capacity credibility coefficient, and the adjustable capacity, capacity constraints are constructed. The upper limit of the maximum allowable response time is determined based on the penetration rate interval identifier and the preset maximum allowable response time mapping relationship. Time constraints are then constructed by combining the clearing capacity decision variables and response time parameters. A hierarchical clearing optimization model is constructed based on multi-level supply and demand balance constraints, capacity constraints, time constraints, and a preset objective function for minimizing operating costs. The solution scheduling module is used to solve the hierarchical clearing optimization model to obtain the target clearing capacity corresponding to each candidate reserve resource among multiple candidate reserve resources; Based on the target clearing capacity, a joint clearing schedule is performed on multiple candidate reserve resources.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the power system graded reserve joint clearing method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device where the storage medium is located to perform the power system graded reserve joint clearing method as described in any one of claims 1 to 7.

Images