Power transmission network power distribution design optimization method and system based on particle swarm optimization
By constructing a working condition vector and a knowledge base of optimal solutions, and combining it with the particle swarm optimization algorithm, the problem of rapid response in the design phase of DC transmission networks was solved, achieving efficient power allocation under emergency conditions and improving the stability and security of the power grid.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- JIANGYIN CHANGYI GRP CO LTD
- Filing Date
- 2026-04-09
- Publication Date
- 2026-05-08
AI Technical Summary
Existing particle swarm optimization algorithms cannot quickly respond to power fluctuations under emergency conditions during the design phase of DC transmission networks, leading to grid stability and security issues and failing to meet the requirements for rapid simulation iteration within 200ms.
By constructing a working condition vector and combining it with the particle swarm optimization algorithm, a hot start is performed using a best solution knowledge base to reduce the invalid search space and improve the algorithm response speed. The system operating status is quantified by the grid state deviation and power trend characteristics to achieve rapid convergence.
In emergency situations, it significantly improves the optimization timeliness and operational stability of the power transmission network, ensuring that the power grid provides high-quality optimization decisions in a very short time and preventing power imbalance and malfunction of protection devices.
Smart Images
Figure CN121997777A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of electrical digital data processing technology, and in particular to a method and system for optimizing power allocation design in power transmission networks based on particle swarm optimization. Background Technology
[0002] With the transformation of the global energy structure and the rapid development of large-scale power transmission networks, DC power optimization technology has become a core component in ensuring the stable operation of power systems after their completion during the design phase of DC transmission networks. Particle Swarm Optimization (PSO), as a classic swarm intelligence algorithm, is applied to the field of power allocation scheme optimization design and simulation verification for power system network nodes due to its strong global search capability and ease of implementation.
[0003] In current DC transmission network design and simulation verification, large-scale transmission networks typically integrate a high proportion of intermittent renewable energy sources such as photovoltaics and wind power. This makes the system operation scenarios that need to be covered in the design phase complex and highly dynamic. For example, the design phase needs to simulate and verify emergency conditions such as sudden cloud cover causing a reduction in photovoltaic power plant output, or unexpected shutdown of wind farms due to sudden faults. In these situations, the transmission network will face severe power fluctuations and frequency deviations. Under these conditions, the control system must quickly and dynamically redistribute the power across the entire network within a short period (usually within the golden response window of 200ms) to maintain the stability of node voltage and power. This is a core operational condition adaptability requirement that must be verified in the transmission network design phase.
[0004] However, existing technologies have limitations in addressing the complex scenarios of multi-condition simulation optimization during the design phase of DC transmission networks. In emergency scenarios involving sudden changes in renewable energy output or system load, the use of traditional particle swarm optimization (PSO) algorithms for power allocation scheme design and simulation optimization suffers from significant time wastage due to the algorithm's initial blind distribution and lack of adaptive mechanisms for actual operating conditions. This results in the output of the power optimization scheme lagging far behind the 200ms short-time response requirement of the power system, failing to meet the efficiency demands of rapid simulation iteration under multiple operating conditions during the transmission network design phase. This "decision lag" in emergency scenarios not only fails to promptly mitigate severe grid fluctuations but also easily triggers a chain reaction of power imbalance across the entire network, even causing malfunctions in relay protection devices. This directly threatens the safety and stable operation of equipment in the entire large-scale DC network and cannot guarantee the reliability of the final designed transmission network scheme. Summary of the Invention
[0005] To improve the timeliness of power distribution optimization response and enhance the stability of equipment operation within the transmission network, this application provides a power allocation design optimization method and system for the transmission network based on the particle swarm optimization algorithm.
[0006] Firstly, this application provides a power allocation design optimization method for power transmission networks based on particle swarm optimization, employing the following technical solution: The power allocation design optimization method for power transmission networks based on particle swarm optimization includes: obtaining the simulated time-series power of each node during the operation of the power transmission network under simulated conditions, and simultaneously obtaining the frequency deviation of each node in the power transmission network. The operating condition vector is constructed based on the simulated time-series power and frequency deviation. The operating condition vector includes at least the grid state deviation, which characterizes the fluctuation amplitude of the overall simulated time-series power of the transmission network, and the power trend feature, which characterizes the overall simulated time-series power trend of the transmission network. The power distribution scheme of the power transmission network is optimized by retrieving the operating condition vector that is closest to the current operating condition vector and its corresponding optimal power allocation scheme from the pre-stored optimal solution knowledge base, and the power allocation scheme is used as the center of the initial particle swarm.
[0007] The simulation time-series power and frequency deviation of each node are obtained. The simulation time-series power is used to characterize the load-supply relationship of the nodes, and the frequency deviation is used to reflect the degree of energy imbalance of the overall system. Furthermore, by constructing an operating condition vector, the complex power grid operating state is expressed through power grid state deviation and power trend characteristics. The power grid state deviation is used to quantify the magnitude of the current system's deviation from steady state, and the power trend characteristics are used to characterize the dynamic trend of power changes. This enables simultaneous characterization of drastic fluctuations and potential changes, significantly improving the ability to identify complex operating conditions.
[0008] Building upon this foundation, by retrieving similar operating conditions from the optimal solution knowledge base and using them as the initial center for the particle swarm optimization (PSO) algorithm, historical experience is introduced into the current optimization process. This ensures that the initial distribution of the particle swarm is no longer random but concentrated in the neighborhood of historical high-quality solutions, reducing the ineffective search space. Simultaneously, combined with PSO optimization, the system performs a refined search within this high-quality region, achieving a balance between global and local optimization. The synergistic effect of these technical features enables the PSO algorithm to converge quickly under sudden power fluctuations or frequency anomalies, shortening the decision-making time. This effectively solves the "response lag" problem caused by the random initialization of the PSO algorithm in existing technologies, enhancing power grid stability and security.
[0009] Optionally, the calculation steps for the power grid state deviation include: calculating the local fluctuation intensity of each node in the transmission network; taking the average of the local fluctuation intensity of each node as the global fluctuation index; and taking the product of the global fluctuation index and the normalized value of the frequency deviation as the power grid state deviation.
[0010] The local fluctuation intensity quantifies the power fluctuation range, accurately reflecting the short-term fluctuation characteristics of each node. The average local fluctuation intensity of each node yields the global fluctuation index, achieving a unified measurement of the overall fluctuation level of the entire network and providing a consistent evaluation standard for networks of different sizes. Furthermore, the global fluctuation index is multiplied and fused with the normalized value of frequency deviation, resulting in a significant increase in the deviation of the power grid state when power fluctuations and frequency anomalies occur simultaneously. This enhances the sensitivity to dangerous operating conditions and reduces misjudgments.
[0011] Optionally, the local fluctuation intensity is positively correlated with the fluctuation range of the simulated time-series power of the corresponding node and the importance coefficient of the node. The importance coefficient is proportional to both the electrical connectivity and global topological centrality of the corresponding node. The electrical connectivity is the sum of the admittances of all branches connected to the node, and the global topological centrality is the reciprocal of the average electrical distance from the node to all other nodes in the power grid.
[0012] An importance coefficient is introduced into the local fluctuation intensity to achieve differentiated modeling of key nodes. Among them, electrical connectivity reflects the connection strength between the node and the network, which can reflect its role in power transmission; global topology centrality reflects the positional importance of the node in the overall structure. Together, they constitute the importance evaluation of the node.
[0013] Optionally, the method for obtaining power trend characteristics includes: for any node in the transmission network, calculating the sum of the absolute values of the simulation time-series power differences between adjacent sampling points within a preset time window to obtain a local trend index; and using the average value of the local trend indices of each node as the power trend characteristic.
[0014] The differences in simulated time-series power between adjacent sampling points are calculated and summed to obtain the trend of nodal power change. If the power shows an increasing or decreasing trend within the preset time window, the difference in simulated time-series power between adjacent moments increases, and thus the power trend characteristic increases. If there is no changing trend in power within the preset time window, the difference in simulated time-series power between each moment approaches 0, and thus the power trend characteristic decreases.
[0015] Optionally, the construction method of the optimal solution knowledge base includes: dividing the feature space of the operating condition vector into grids to generate an initial grid point set; performing particle swarm optimization algorithm on each grid point to find the optimal solution, saving the obtained operating condition vector and the optimal power allocation scheme to obtain the initial knowledge base; obtaining the operating condition vector during the historical operation of the transmission network, adjusting the grid division interval in the feature space according to the distribution of the historical operating condition vector, performing particle swarm optimization algorithm on the newly generated grid points to calculate their optimal power allocation scheme and saving it to obtain the optimal solution knowledge base.
[0016] Grid partitioning discretizes the continuous operating space, facilitating systematic coverage of different operating states of the power transmission network. Secondly, the optimal solution, i.e. the optimal power allocation scheme, is pre-calculated for each grid point using an offline particle swarm optimization algorithm, transferring high computational load to the offline stage and reducing computational burden. Thirdly, the grid interval is adjusted by historical data distribution, making the high-frequency operating area more finely divided and the low-frequency area more coarse-grained, thereby controlling storage and computational complexity while ensuring accuracy.
[0017] Optionally, the system may retrieve the operating condition vector that is closest to the current operating condition vector and its corresponding optimal power allocation scheme from a pre-stored optimal solution knowledge base, including: calculating the Euclidean distance between the current operating condition vector and all historical operating condition vectors in the knowledge base.
[0018] Euclidean distance is used as the working condition matching metric. By calculating the distance between the current working condition situation vector and the working condition situation vector in the knowledge base, the most similar working condition can be quickly retrieved.
[0019] Optionally, it also includes: obtaining the fitness value of the optimal power allocation scheme obtained from the current design simulation; when the fitness value is better than a preset effective threshold, and the minimum distance between the current operating condition vector and all operating condition vectors in the knowledge base is greater than a preset threshold, storing the current operating condition vector and its corresponding optimal power allocation scheme into the optimal solution knowledge base.
[0020] The knowledge base is dynamically updated through a dual judgment mechanism that uses both a fitness threshold and a distance-based preset threshold. The fitness threshold ensures that only high-quality solutions are included, preventing low-quality data from polluting the knowledge base.
[0021] Optionally, it also includes: forcibly terminating the iterative process of the particle swarm optimization algorithm within a preset response time; and outputting the optimal power allocation scheme.
[0022] This mechanism avoids the problem of response delay caused by excessive pursuit of the optimal solution, and can prioritize the stability of the system in actual power grids.
[0023] Optionally, the grid partitioning interval in the feature space is adjusted according to the distribution of historical operating condition vectors, including: clustering historical operating condition vectors to obtain multiple clusters; for any region where a cluster is located, adjusting the grid partitioning interval of the corresponding region according to the intra-cluster sample density of the cluster, wherein the intra-cluster sample density is negatively correlated with the grid partitioning interval.
[0024] Secondly, this application provides a power allocation design optimization system for power transmission networks based on particle swarm optimization, employing the following technical solution: A power grid power allocation design optimization system based on particle swarm optimization algorithm includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the power grid power allocation design optimization method based on particle swarm optimization algorithm described above.
[0025] The above-mentioned power distribution design optimization method for power transmission networks based on particle swarm optimization algorithm is generated into a computer program and stored in memory for loading and execution by the processor. Thus, a system is built based on the memory and processor for convenient use.
[0026] This application has the following technical effects: By extracting deviation and trend features to construct operating condition vectors, the system's operating environment is quantified, and a knowledge base of optimal solutions is pre-established. Under emergency conditions, the system accurately matches the closest operating condition vector and extracts the corresponding optimal power allocation scheme to perform a warm start of the algorithm, reducing the algorithm's invalid search space, improving the speed of outputting the optimal power allocation scheme, and enhancing the timeliness and operational stability of power transmission network optimization. Attached Figure Description
[0027] Figure 1 This is a schematic diagram of the power allocation design optimization method for power transmission networks based on particle swarm optimization in this application embodiment.
[0028] Figure 2 This is a schematic diagram illustrating the construction method of the optimal solution knowledge base in the power allocation design optimization method for power transmission networks based on particle swarm optimization in this application embodiment. Detailed Implementation
[0029] This application discloses a power allocation design optimization method for power transmission networks based on particle swarm optimization. First, it obtains the simulated time-series power and frequency deviations of the transmission network, and extracts the grid state deviation and power trend features to construct an operating condition vector. Based on this operating condition vector, it performs distance matching with a best solution knowledge base to achieve… The algorithm undergoes a warm-start initialization. Further enhanced by a forced termination mechanism, it outputs the optimal power allocation scheme while meeting system response time limits, and dynamically updates the optimal solution knowledge base based on the results. This scheme improves the algorithm's optimization efficiency and response speed under emergency conditions, ensuring high-quality optimization decisions are provided in a very short time, thereby effectively guaranteeing the safe and stable operation of the transmission network under complex conditions.
[0030] Reference Figure 1 The power allocation design optimization method for power transmission networks based on particle swarm optimization includes steps S1-S3.
[0031] S1: Obtain the simulated time-series power of each node during the operation of the power transmission network under simulated conditions, and simultaneously obtain the frequency deviation of each node in the power transmission network.
[0032] In large-scale power transmission networks, high-density monitoring of the overall network operation status is required to accurately capture the dynamic changes brought about by the integration of renewable energy.
[0033] For example, with The sampling period is used to obtain the simulated time-series power of each node in the network.
[0034] Simultaneously, the simulated timing frequencies of the AC busbars of converter stations at each node of the entire transmission network are acquired synchronously, and the frequency deviation is obtained by subtracting the grid's rated frequency from the actual frequency. In this embodiment, the rated frequency is 50Hz. After acquisition, the acquired data sequences are time-aligned so that the sampling points of the simulated timing frequencies strictly correspond to the end points of the preset power time window, thereby acquiring all the data in the transmission network. Simulation timing power and frequency deviation of each node.
[0035] S2: Construct a working condition situation vector based on the simulation time-series power and frequency deviation. The working condition situation vector shall include at least the grid state deviation, which characterizes the fluctuation amplitude of the overall simulation time-series power of the transmission network, and the power trend feature, which characterizes the overall simulation time-series power trend of the transmission network.
[0036] The intermittent output of renewable energy and sudden changes in load can lead to rapid changes in system state. Traditional optimization algorithms with fixed parameters cannot simultaneously ensure computational accuracy under normal operating conditions and convergence speed under emergency conditions. Therefore, this embodiment constructs multi-dimensional features as a quantitative evaluation basis to achieve adaptive matching of the algorithm.
[0037] Set the preset time window length to It can not only cover the initial characteristics of the system's transient process, but also avoid the judgment bias caused by sampling lag, providing high-precision data support for the identification of subsequent complex working conditions.
[0038] The calculation steps for grid state deviation include: calculating the local fluctuation intensity of each node in the transmission network; taking the average local fluctuation intensity of each node as the global fluctuation index; and taking the product of the global fluctuation index and the normalized value of the frequency deviation as the grid state deviation.
[0039] In one embodiment, the formula for calculating the intensity of local fluctuations can be expressed as: In the formula, Indicates the first in the network The local fluctuation intensity of each node; Indicates the first The maximum power of each node within the preset time window corresponding to the current moment; Indicates the first The minimum power of each node within a preset time window at the current moment; Indicates the first The rated maximum power of each node; Indicates the first The importance coefficient of a node is used to reflect the criticality of that node in the overall network topology.
[0040] Specifically, the formula for calculating the importance coefficient can be expressed as: In the formula, Indicates the first The importance coefficient of each node; Represents a node Electrical connectivity (i.e., with nodes) (Total number of connected branch admittances). Represents the global topological centrality; this value is taken from the node's... The reciprocal of the average electrical distance to all other nodes in the power grid (previous technology, calculated based on the node impedance matrix, which is also prior art). This ensures that the importance coefficient of critical nodes located at topological hubs approaches 1.
[0041] In another embodiment, the local fluctuation intensity can also be calculated using the following formula: In the formula, Indicates the first in the network The local fluctuation intensity of each node; Indicates the first The maximum power of each node within the preset time window corresponding to the current moment; Indicates the first The minimum power of each node within a preset time window at the current moment; Indicates the first The rated maximum power of each node.
[0042] This formula omits the importance coefficient, reduces the amount of calculation, and can improve the system's computational efficiency. In specific implementation, those skilled in the art can choose according to the actual situation.
[0043] Alternatively, in some embodiments, the standard deviation of the simulated time-series power at each moment within a preset time window can be used as the local fluctuation intensity.
[0044] The average instantaneous fluctuation intensity of each node in the transmission network is used as the global fluctuation index, and the product of the global fluctuation index and the normalized result of the current frequency deviation is used as the current grid state deviation of the transmission network.
[0045] It is mainly used to measure the degree to which the overall power grid state deviates from steady state. The formula for calculating the degree of power grid state deviation can be expressed as: ; In the formula, This represents the deviation of the power grid state of the transmission network at the current moment, and its value range is distributed in the following ranges: arrive between; This represents the global fluctuation index of the power transmission network; This indicates the frequency deviation at the end of the current preset time window; This represents the maximum frequency deviation recorded in historical operating data, and is mainly used for normalization of the numerator.
[0046] Based on the above formula structure, when the power extreme value difference among the highly important nodes in the network increases and the grid frequency deviation increases sharply, the final calculated grid state deviation will be... The value will increase significantly. This increase in the eigenvalue directly reflects that the system is in the early stages of instability caused by energy cut-off or load surge, thus providing a clear quantitative threshold for triggering the accelerated response of subsequent optimization algorithms.
[0047] For the power trend characteristics of the transmission network, for any node in the transmission network, the local trend index is calculated based on the adjacent differences of the simulated time series power of each adjacent time in the preset time window corresponding to the current time of the node; the mean of the local trend index of each node in the transmission network is used as the power trend characteristic.
[0048] The formula for calculating the local trend index can be expressed as: In the formula, Indicates the first in the power transmission network Local trend index of each node; Indicates the number of sampling points within the preset time window; Indicates the first The nth node in the window Simulated timing power at each sampling point; Indicates the first The nth node in the window Simulated timing power at each sampling point.
[0049] As the formula shows, by taking the absolute value of the power difference between adjacent sampling points and then normalizing and summing them, instantaneous power fluctuations can be effectively captured. In practical applications, even if the system frequency deviation has not yet accumulated to cause a deviation in the grid state... If the power trend remains low, A high value clearly indicates that the current DC network is in a high-frequency oscillation state, which is highly likely to evolve into a global instability condition. This should be considered in conjunction with the aforementioned grid state deviation. Power trend characteristics This yields a condition vector that accurately describes the simulation conditions of the system design.
[0050] The operating condition vector can be represented as ,in, Indicates the deviation of the power grid state. This indicates the power trend characteristics.
[0051] S3: Retrieve the operating condition vector closest to the current operating condition vector and its corresponding optimal power allocation scheme from the pre-stored optimal solution knowledge base, and use the particle swarm optimization algorithm to optimize the power of the transmission network, wherein the optimal power allocation scheme serves as the center of the initial particle swarm.
[0052] This step first explains the construction steps of the pre-stored optimal solution knowledge base.
[0053] The steps for constructing the pre-stored optimal solution knowledge base include: S301-step S302.
[0054] S301: Perform grid division in the feature space of the working condition situation vector to generate an initial grid point set; perform particle swarm optimization algorithm to optimize each grid point, and save the obtained working condition situation vector and the optimal power allocation scheme to obtain the initial knowledge base.
[0055] As can be seen from step S2, the operating condition vector used to evaluate the power grid state includes the power grid state deviation degree used to assess the power grid fluctuation amplitude and the power trend characteristic used to assess the power grid power change trend, and both are... arrive Continuous variables within a space.
[0056] In this step, an initial grid spacing is set. In this embodiment, the initial grid spacing is 0.2. arrive Within the space, grid points are uniformly generated according to the initial grid division interval to form an initial grid point set. Then, offline optimization is performed on the grid points in the initial grid point set. During the offline optimization process, a fitness function is constructed to evaluate the quality of the power allocation scheme. The formula for calculating the fitness function can be expressed as: ; In the formula, Indicates the fitness value; This indicates the total available power of the entire network under the current operating conditions; Indicates the first power allocation scheme under the current power allocation scheme. Power allocation values for each node; This represents the baseline power, which is the sum of the maximum rated power of all nodes in the network. Indicates the first The tendency coefficient of each node is directly taken as the local fluctuation intensity of that node; Indicates the first Target power setting for each node.
[0057] The formula uses the absolute deviation between the actual allocated total power and the total available power of the system as a global penalty term, while multiplying the local deviation of each node by its corresponding tendency coefficient as a local constraint term. When a critical node experiences drastic fluctuations, its tendency coefficient increases, forcing the optimization algorithm to prioritize reducing the allocation deviation of that node during the optimization process to lower the overall fitness value. This design guides the algorithm to automatically allocate adjustment resources to key nodes that urgently need stability.
[0058] After offline optimization, each grid point in the initial grid set is associated with and recorded as an optimal power allocation scheme, i.e., the solution vector output by the particle swarm optimization algorithm. The optimal power allocation scheme is composed of the optimal power of each node in the transmission network. The records of offline training for each grid point constitute the initial knowledge base. This can also be understood as the optimal solution knowledge base storing records used for offline training based on the particle swarm optimization algorithm. Each record contains a binary tuple: , This is the working condition status vector; This represents the corresponding optimal power allocation scheme, i.e., the solution vector.
[0059] S302: Obtain the operating condition vector of the power transmission network during its historical operation, adjust the grid division interval in the feature space according to the distribution of the historical operating condition vector, execute the particle swarm optimization algorithm on the newly generated grid points to calculate its optimal power allocation scheme and save it to obtain the optimal solution knowledge base.
[0060] The historical working condition vectors are clustered to obtain multiple clusters. For any region containing a cluster, the grid division interval of the corresponding region is adjusted according to the sample density within that cluster.
[0061] In this embodiment, the K-Means clustering algorithm is used to cluster the operating condition vectors during the historical normal operation of the power transmission network. During the clustering process, The value can be adaptively selected using the elbow method, which is a conventional technique for those skilled in the art and will not be elaborated upon here.
[0062] Subsequently, the cluster center and intra-cluster sample density of each cluster are obtained. The formula for calculating the intra-cluster sample density can be expressed as: ; In the formula, Indicates the first The intra-cluster sample density of each cluster; Indicates the first The number of samples within each cluster; Indicates the first The coverage radius of each cluster is the 95th percentile of the distance from the sample within the cluster to the cluster center.
[0063] The grid spacing of the corresponding region is adjusted based on the sample density within the cluster. The corresponding calculation formula can be expressed as: ; In the formula, Indicates the first Local partitioning intervals of individual clusters; Indicates the initial partition interval; This represents the standard normalization function; Indicates the first The mean fitness of the initial grid points in each cluster; Indicates the first The intra-cluster sample density of each cluster.
[0064] When the mean fitness of a cluster is high, it indicates poor solution quality in that region; simultaneously, a high sample density within that cluster suggests it belongs to a high-frequency, frequent operating condition. Therefore, the calculated local partitioning interval... This will reduce the size accordingly. It is understood that in this embodiment, the local partitioning interval is calculated based on two dimensions (intra-cluster sample density and mean fitness). In other embodiments, to improve computational efficiency, the local partitioning interval can also be calculated based solely on the intra-cluster sample density.
[0065] At this point, a denser grid of points is generated within the characteristic region, and offline optimization is applied to the newly generated grid points, thereby increasing the number of records for this region in the optimal solution knowledge base and providing sufficient reference data for complex and variable operating conditions. The operating condition vectors of all acquired grid points and their corresponding optimal power allocation schemes are saved to form the optimal solution knowledge base.
[0066] During the design and simulation phase, the current working condition vector is acquired, and its standard Euclidean distance with all recorded working condition vectors in the optimal solution knowledge base is calculated. The working condition vector with the smallest distance is selected, and its corresponding solution vector is extracted and directly used as the initial particle swarm center for this particle swarm optimization algorithm. Random perturbations are introduced near this center to generate the initial population.
[0067] This hot-start method can directly lock the initial search range of the particle swarm within the neighborhood of historical high-quality solutions, fundamentally overcoming the defect of excessive invalid iterations in the early stage caused by blindly scattering points randomly.
[0068] To ensure rapid and stable convergence of the algorithm after a warm start, implementation limits for the core hyperparameters of the particle swarm optimization algorithm are set. Specifically, the population size is set to 80, the individual learning factor and the social learning factor are both set to 2.0, and the inertia weight is set to decrease linearly from 0.9 to 0.4 with the number of iterations. A larger initial weight helps the algorithm escape local optima, while linear decay ensures finer exploration of the neighborhood of the recommended solution in the knowledge base.
[0069] A time-constrained forced termination mechanism is set up during the iteration process of the particle swarm optimization algorithm to ensure that the power allocation optimization meets the power system response requirements.
[0070] When the algorithm iteration is started, a hardware timer is started simultaneously. When the hardware timer reaches the preset duration, regardless of whether the current population convergence state meets the accuracy requirements, the iteration is forcibly terminated and the globally optimal solution vector is output to the power transmission network simulation control system.
[0071] The preset duration here is no more than 180ms. Since the golden window for actual network fault response is only 200ms, setting a forced termination time of 180ms allows for a sufficient 20ms for the execution and output of communication and adjustment commands from underlying devices. When the system faces wind turbine shutdown or a sudden drop in photovoltaic output, this flexible cutoff mechanism sacrifices a very small amount of optimization accuracy, but in exchange for the overall safety of the system, preventing malfunctions of protection devices caused by delayed optimization commands.
[0072] Finally, the fitness values of all records in the optimal solution knowledge base are statistically analyzed, and the upper quartile is selected as the effective fitness threshold. If the fitness of the optimal power allocation scheme output by the current time series is less than this effective fitness threshold, the fitness of the optimal power allocation scheme output by the current time series is considered to be within the excellent range. Further, the knowledge base is searched for similar records whose distance to the current operating condition vector is less than a preset distance (0.15). If such a record exists, the one with the smaller fitness value is retained; otherwise, the current operating condition vector and its corresponding optimal power allocation scheme are added to the knowledge base as new records. As the system continues to operate and equipment performance improves, this feedback loop logic enables the optimal solution knowledge base to self-correct and dynamically expand, thereby maintaining long-term optimization accuracy.
[0073] This application also discloses a power grid power allocation design optimization system based on particle swarm optimization algorithm, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the power grid power allocation design optimization method based on particle swarm optimization algorithm according to this application is implemented.
[0074] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.
[0075] In this application, the aforementioned memory can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store required information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device.
[0076] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. A power allocation design optimization method for power transmission networks based on particle swarm optimization, characterized in that, The simulation time-series power of each node in the power transmission network is obtained under the simulation conditions, and the frequency deviation of each node in the power transmission network is obtained synchronously. The operating condition vector is constructed based on the simulated time-series power and frequency deviation. The operating condition vector includes at least the grid state deviation, which characterizes the fluctuation amplitude of the overall simulated time-series power of the transmission network, and the power trend feature, which characterizes the overall simulated time-series power trend of the transmission network. The power distribution scheme of the power transmission network is optimized by retrieving the operating condition vector that is closest to the current operating condition vector and its corresponding optimal power allocation scheme from the pre-stored optimal solution knowledge base, and the power allocation scheme is used as the center of the initial particle swarm.
2. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, The calculation steps for grid state deviation include: calculating the local fluctuation intensity of each node in the transmission network; taking the average local fluctuation intensity of each node as the global fluctuation index; and taking the product of the global fluctuation index and the normalized value of the frequency deviation as the grid state deviation.
3. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 2, characterized in that, The intensity of the local fluctuation is positively correlated with the fluctuation range of the simulated time-series power of the corresponding node and the importance coefficient of the node. The importance coefficient is proportional to the electrical connectivity and global topological centrality of the corresponding node. The electrical connectivity is the sum of the admittances of all branches connected to the node, and the global topological centrality is the reciprocal of the average electrical distance from the node to all other nodes in the power grid.
4. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, The method for obtaining power trend characteristics includes: for any node in the transmission network, calculating the sum of the absolute values of the simulation time-series power differences between adjacent sampling points within a preset time window to obtain the local trend index; and taking the average value of the local trend index of each node as the power trend characteristic.
5. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, The method for constructing the optimal solution knowledge base includes: dividing the feature space of the operating condition vector into grids to generate an initial grid point set; performing particle swarm optimization on each grid point to find the optimal solution, saving the obtained operating condition vector and the optimal power allocation scheme to obtain the initial knowledge base; obtaining the operating condition vector during the historical operation of the transmission network, adjusting the grid division interval in the feature space according to the distribution of the historical operating condition vector, performing particle swarm optimization on the newly generated grid points to calculate their optimal power allocation scheme and saving it to obtain the optimal solution knowledge base.
6. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, Retrieve the operating condition vector that is closest to the current operating condition vector and its corresponding optimal power allocation scheme from the pre-stored optimal solution knowledge base, including: calculating the Euclidean distance between the current operating condition vector and all historical operating condition vectors in the knowledge base.
7. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, Also includes: Obtain the fitness value of the optimal power allocation scheme obtained from the current design simulation run; When the fitness value is better than the preset effective threshold, and the minimum distance between the current operating condition vector and all operating condition vectors in the knowledge base is greater than the preset threshold, the current operating condition vector and its corresponding optimal power allocation scheme are stored in the optimal solution knowledge base.
8. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 1, characterized in that, Also includes: The iteration process of the particle swarm optimization algorithm is forcibly terminated within a preset response time. Output the optimal power allocation scheme.
9. The power allocation design optimization method for power transmission networks based on particle swarm optimization algorithm according to claim 5, characterized in that, Adjusting the grid partitioning interval in the feature space based on the distribution of historical operating condition vectors includes: clustering historical operating condition vectors to obtain multiple clusters; for any region containing a cluster, adjusting the grid partitioning interval of the corresponding region based on the intra-cluster sample density of that cluster, wherein the intra-cluster sample density is negatively correlated with the grid partitioning interval.
10. A power allocation design optimization system for power transmission networks based on particle swarm optimization, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the power allocation design optimization method for power transmission networks based on the particle swarm optimization algorithm according to any one of claims 1-9.