A power distribution network voltage stability optimization method
By acquiring and analyzing real-time data, and combining support vector machines and convolutional neural network models, the allocation of distribution network resources is optimized, which solves the problem of voltage instability in the distribution network under instantaneous load fluctuations, and realizes rapid response to local areas and improves power supply reliability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID SHANDONG ELECTRIC POWER CO LAIXI CITY POWER SUPPLY CO
- Filing Date
- 2026-03-25
- Publication Date
- 2026-07-21
AI Technical Summary
Existing power distribution network management methods are unable to cope with instantaneous load fluctuations in complex operating environments, leading to voltage instability and the risk of local power outages, and lack the ability to respond quickly to short-term, dynamic changes in local areas.
By collecting electricity consumption data and voltage signals in real time through sensor networks, the system uses support vector machine algorithm to classify fluctuation types and severity, combines historical data for trend analysis, triggers voltage instability warnings, and uses convolutional neural network model to predict short-term load demand, determines energy regulation priority sequence, allocates backup power and energy storage equipment resources, and iteratively optimizes resource paths to improve power supply reliability.
It significantly improves the voltage stability and response efficiency of the distribution network, reduces the risk of voltage instability, and ensures an efficient and reliable power supply.
Smart Images
Figure CN122436982A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system control, and more specifically to a method for optimizing voltage stability in a distribution network. Background Technology
[0002] In modern power systems, the distribution network, as a crucial link connecting generation and consumption, plays an irreplaceable role in ensuring electricity supply for society's production and daily life. Especially given the increasing demand for electricity and intensified load fluctuations, effectively balancing grid load and improving power supply reliability have become critical issues that the power sector urgently needs to address. The operation of the distribution network directly affects the electricity experience of countless households; therefore, researching ways to optimize its performance is particularly important.
[0003] However, current methods commonly used in distribution network management often struggle to adapt to complex operating environments, especially when facing sudden load changes and regional peak electricity demand. Existing methods frequently fall short. These methods are primarily based on overall planning or long-term dispatching, lacking the ability to quickly respond to short-term, dynamic changes within local areas. This limitation makes the power grid prone to stability issues when dealing with instantaneous fluctuations, thus affecting the power quality for users.
[0004] Instantaneous load fluctuations in power distribution networks pose a primary challenge. Due to the unpredictability of electricity demand, short-term, drastic load changes may occur in certain periods or areas. For example, when large equipment is started up in an industrial area, a sudden surge in electricity consumption can lead to a sharp increase in pressure on the local power grid. This volatility further exacerbates the difficulty of peak load management, because failure to balance such sudden electricity demand in a timely manner may lead to voltage instability or even local power outages. Instantaneous fluctuations and peak loads are closely related; the former is often the direct cause of the latter, while the latter poses a greater threat to the overall stability of the power grid.
[0005] This issue not only concerns technological innovation but also directly affects users' actual experience in daily electricity use, and urgently requires in-depth research and solutions. Summary of the Invention
[0006] The purpose of this invention is to solve the above-mentioned problems and provide a method for optimizing voltage stability in power distribution networks.
[0007] The technical solution adopted by this invention to solve its technical problem is: A method for optimizing voltage stability in a power distribution network includes the following steps: S101 collects electricity consumption data and voltage signals in local areas of the power distribution network in real time through a sensor network, processes these data to identify instantaneous fluctuation characteristics, and obtains instantaneous load fluctuation sequences.
[0008] Based on the obtained instantaneous load fluctuation sequence, S102 uses the support vector machine algorithm to classify the fluctuation type and severity, and determine the potential location and time point of peak increase.
[0009] S103 obtains the determined peak increase location and time point, integrates historical load data for trend analysis, and determines that if the peak increase exceeds the preset threshold, it will trigger a voltage instability warning and obtain a voltage instability risk assessment result.
[0010] Based on the obtained voltage instability risk assessment results, S104 extracts the dynamic change parameters of the local area, predicts short-term load demand through a convolutional neural network model, and determines the priority sequence of energy regulation.
[0011] S105 selects a high-priority region from the determined energy regulation priority sequence, calculates the difference between available energy reserves and demand, and obtains a load balance adjustment scheme.
[0012] S106 allocates backup power and energy storage equipment resources according to the obtained load balancing adjustment scheme. If the remaining fluctuation response is insufficient after allocation, the resource path is iteratively optimized to determine the final power supply reliability improvement configuration.
[0013] S107 obtains the final power supply reliability improvement configuration and transmits it to the actuator in the control center in real time. It monitors the voltage signal feedback after execution and obtains optimized performance verification data.
[0014] Further, step S101 includes: The power consumption data and voltage signals of local areas of the power distribution network are collected in real time through sensor networks to obtain raw data streams; After preprocessing the original data stream, a purified data signal is obtained, and time series segments related to instantaneous fluctuations in the purified data signal are extracted to determine the time point and amplitude range of the fluctuations.
[0015] Further, step S102 includes: Instantaneous load fluctuation sequence data is obtained from the monitoring system, and noise and outliers are removed through preprocessing to obtain cleaned load sequence data; For the cleaned load sequence data, the support vector machine algorithm is used to classify the fluctuation type, determine whether the fluctuation is periodic or sudden, and obtain the classification result of the fluctuation type. Based on the classified fluctuation type results, the severity of the fluctuation type is analyzed. If the fluctuation amplitude exceeds the preset threshold, it is marked as high severity, and severity assessment data is obtained. Using the severity assessment data, anomalies with increased peak values are identified, and time series analysis tools are used to locate the potential locations of the increased peak values to obtain the spatial distribution information of the anomalies. Based on the spatial distribution information of the anomalies and combined with time series data, the specific time nodes of the peak increase are determined, and the precise labeled data of the time nodes are obtained. By using the precisely labeled data of the time nodes and the spatial distribution information, a comprehensive analysis result of load fluctuation anomalies is generated to determine the final peak increase location and time point.
[0016] Further, step S103 includes: Historical load records are retrieved from a pre-established load data repository, load change data within a specified time range is compiled, and daily peak points and corresponding time nodes are determined. Based on the daily peak points and time nodes, calculate the peak increment within adjacent time periods, analyze the peak change trend, and obtain the continuity characteristics of peak change; Based on the continuous characteristics of the peak value changes, identify the abnormal locations and specific time points of the peak value increments, record the distribution of the abnormal increments, and determine potential risk points. If the peak increment exceeds a preset threshold, the abnormal location and time point will be marked as a high-risk area, triggering a preliminary warning signal. For the high-risk areas, a support vector machine model is used to classify and predict the probability of voltage instability by combining historical load data and voltage instability records. The corresponding risk assessment level is generated based on the probability of voltage instability occurring.
[0017] Further, step S104 includes: Voltage instability data for local areas is obtained from the risk assessment results. Automated tools are used to perform layered processing on the voltage instability data to obtain a set of dynamically changing parameters within the area. Based on the set of dynamically changing parameters, a convolutional neural network model is used to predict short-term load change trends and determine the load fluctuation range for each region. For the load fluctuation range, if the fluctuation of a certain area exceeds a preset threshold, the area is marked as a high-risk object, and the energy regulation demand sequence of the high-risk object is obtained; By combining the energy regulation demand sequence with regional division data, the load transferability between regions is determined, and a preliminary priority order list is obtained. Based on the preliminary priority list, the matching degree between the dynamic change parameters of each region and the load forecast results is obtained, and the final energy regulation priority sequence is determined.
[0018] Further, step S105 includes: Priority data is obtained from a pre-established energy regulation database, and high-priority areas are initially screened to determine a list of target areas; Based on the target area list, obtain the available energy and energy reserve data for each area, and determine whether the energy reserves of each area meet the preset threshold by comparing the data. If they are lower than the preset threshold, they are marked as scarce areas. For the aforementioned areas with severe energy shortages, actual demand data is obtained, and statistical tools are used to calculate the difference between the actual demand and the energy reserves to obtain the energy gap value for each area. Based on the energy gap value and the priority data, the energy gap of the high-priority areas is sorted using a linear regression model to determine the emergency allocation order. For the emergency allocation order, obtain the available energy data of the surrounding area. If the available energy in the surrounding area is higher than a preset threshold, generate a cross-regional energy allocation instruction and obtain the allocated energy distribution result. Based on the energy allocation results, the load balance status of each region after the adjustment is calculated, and the load balance status is determined by data comparison to obtain the final adjustment plan.
[0019] Further, step S106 includes: Acquire system load balance data, analyze the resource allocation status in the load balance data, determine the available capacity of backup power and energy storage devices, and generate an initial allocation scheme; Calculate the coverage area of the fluctuation response based on the initial allocation scheme; If the coverage area is lower than a preset threshold, a residual insufficiency determination is triggered to obtain gap data of the fluctuation response; For the aforementioned gap data, an iterative optimization method is used to adjust the resource path, reallocate the output ratio of the backup power supply and the energy storage device, and determine the optimized resource allocation result. Extract key indicators of power supply reliability from the optimized resource allocation results; If the aforementioned key indicators are not met, the resource path will be adjusted again to determine a new allocation combination; Based on the new allocation combination, the matching degree between power supply reliability and the improved configuration is analyzed, and the applicability evaluation value of the final solution is obtained by comparing historical data. Based on the applicability assessment value, resource paths and allocation ratios are locked, and complete data for power supply reliability improvement configuration is generated.
[0020] Further, step S107 includes: Obtain optimized configuration schemes that conform to the current power grid status from a pre-established database of power supply reliability improvement schemes, and determine the appropriate execution parameters; The optimized configuration scheme is sent to the main control system of the control center through a real-time transmission channel, and the timestamp and data integrity status are recorded during the transmission process. Based on the optimized configuration scheme received by the control center, the drive execution device deploys the scheme. If an abnormal signal is detected during the deployment process, the backup scheme is switched and the deployment completion status information is obtained. Based on the deployment completion status information, the voltage signal acquisition module is activated to continuously monitor the voltage signal data output by the execution device and obtain real-time signal fluctuation records. The signal fluctuation records are classified using a support vector machine algorithm to determine whether there are abnormal fluctuations exceeding a preset threshold, and to determine the signal quality assessment result. By combining the signal quality assessment results with historical feedback data, the performance optimization results are analyzed to obtain performance verification data.
[0021] The beneficial effects of this invention are: 1. This invention addresses the unique business scenario of voltage instability risk caused by instantaneous load fluctuations in local areas of a power distribution network. This problem integrates the logical connections between real-time data acquisition and processing, fluctuation type classification, risk assessment, load forecasting, and resource allocation. Specifically, it addresses how to identify potential peak increases, predict short-term demand, and optimize energy regulation to prevent power outages in a dynamic electricity environment. This invention uses a sensor network to collect electricity consumption data and voltage signals in real time. It employs a support vector machine algorithm to classify fluctuation types and severity, integrates historical data for trend analysis to trigger voltage instability early warnings, and utilizes a convolutional neural network model to predict short-term load demand, determine energy regulation priority sequences, calculate the difference between available energy reserves and demand to generate load balancing adjustment schemes, and finally allocate backup power and energy storage resources and iteratively optimize paths. This achieves real-time transmission and execution feedback monitoring of power supply reliability improvement configurations. This method significantly improves the voltage stability and response efficiency of the power distribution network, reduces the probability of risk occurrence, and ensures efficient and reliable power supply. Attached Figure Description
[0022] Figure 1 This is a flowchart illustrating the structure of the present invention. Detailed Implementation
[0023] like Figure 1 As shown, a method for optimizing voltage stability in a distribution network includes the following steps: S101 collects electricity consumption data and voltage signals in local areas of the power distribution network in real time through a sensor network, processes these data to identify instantaneous fluctuation characteristics, and obtains instantaneous load fluctuation sequences.
[0024] Based on the obtained instantaneous load fluctuation sequence, S102 uses the support vector machine algorithm to classify the fluctuation type and severity, and determine the potential location and time point of peak increase.
[0025] S103 obtains the determined peak increase location and time point, integrates historical load data for trend analysis, and determines that if the peak increase exceeds the preset threshold, it will trigger a voltage instability warning and obtain a voltage instability risk assessment result.
[0026] Based on the obtained voltage instability risk assessment results, S104 extracts the dynamic change parameters of the local area, predicts short-term load demand through a convolutional neural network model, and determines the priority sequence of energy regulation.
[0027] S105 selects a high-priority region from the determined energy regulation priority sequence, calculates the difference between available energy reserves and demand, and obtains a load balance adjustment scheme.
[0028] S106 allocates backup power and energy storage equipment resources according to the obtained load balancing adjustment scheme. If the remaining fluctuation response is insufficient after allocation, the resource path is iteratively optimized to determine the final power supply reliability improvement configuration.
[0029] S107 obtains the final power supply reliability improvement configuration and transmits it to the actuator in the control center in real time. It monitors the voltage signal feedback after execution and obtains optimized performance verification data.
[0030] Step S101 includes: acquiring raw data streams by real-time collection of electricity consumption data and voltage signals in a local area of the power distribution network through a sensor network; preprocessing the raw data streams to obtain purified data signals; and extracting time series segments related to instantaneous fluctuations from the purified data signals to determine the time point and amplitude range of the fluctuations.
[0031] When using sensor networks to collect real-time electricity consumption data and voltage signals from a local area of a power distribution network, consider a scenario of power distribution network monitoring in a residential community within a city. Sensors are deployed at the community's substation and key line nodes, collecting voltage and current data once per second to form a raw data stream. This data may contain noise, such as outliers caused by equipment aging or external electromagnetic interference. In the collected raw data stream, voltage values may fluctuate around 220 volts, but occasionally sudden spikes may occur, such as reaching 250 volts, lasting for several milliseconds.
[0032] Preprocessing of the raw data stream can employ filtering techniques to remove noise. For example, median filtering can be used to replace values in the data stream that significantly deviate from the normal range with the median value of adjacent time points, thereby obtaining a purified data signal.
[0033] When extracting time series segments related to instantaneous fluctuations, a threshold can be set, such as marking segments where voltage changes exceed 5 volts and last for less than 1 second as fluctuation events. For example, in a monitoring session, if the voltage is observed to drop sharply from 220 volts to 210 volts at a certain point in time, and then recover after 0.5 seconds, this segment will be extracted, determining the time point of the fluctuation and its amplitude range of 10 volts. When refining the fluctuation sequence, multiple fluctuation segments can be integrated into a simplified time-amplitude sequence, retaining only the key fluctuation start, peak, and end data, reducing redundant information and facilitating subsequent analysis. This refinement process helps improve computational efficiency, especially when processing large-scale data.
[0034] Step S102 includes: acquiring instantaneous load fluctuation sequence data from the monitoring system, removing noise and outliers through preprocessing to obtain cleaned load sequence data; classifying fluctuation types using a support vector machine algorithm for the cleaned load sequence data, determining whether the fluctuations are periodic or sudden, and obtaining the classified fluctuation type results; analyzing the severity of the fluctuation type based on the classified fluctuation type results, marking it as high severity if the fluctuation amplitude exceeds a preset threshold, and obtaining severity assessment data; identifying anomalies with increased peak values using the severity assessment data, locating potential locations of increased peak values using time series analysis tools, and obtaining spatial distribution information of the anomalies; determining the specific time nodes of increased peak values based on the spatial distribution information of the anomalies and the time series data, and obtaining precise labeling data for the time nodes; generating a comprehensive analysis result of load fluctuation anomalies using the precise labeling data of the time nodes and the spatial distribution information, and determining the final location and time of the increased peak value.
[0035] Instantaneous load fluctuation sequence data typically contains changes in electrical load over a short period of time. It may be affected by equipment noise or external interference. These data can be filtered to remove noise and outliers, such as by smoothing the data, to obtain cleaned load sequence data.
[0036] When classifying fluctuation types using the Support Vector Machine (SVM) algorithm on cleaned load sequence data, fluctuations can be categorized into periodic and sudden fluctuations. SVM distinguishes between these fluctuations by constructing a classification boundary and mapping data features to a high-dimensional space. For example, if a load data segment exhibits regular fluctuations throughout the day, it can be classified as a periodic fluctuation; conversely, a sudden jump in load from 200 kW to 800 kW can be classified as a sudden fluctuation.
[0037] When analyzing the severity of fluctuation types, a preset threshold of 500 kilowatts is set. If the amplitude of a sudden fluctuation reaches 700 kilowatts, exceeding the threshold, it is marked as high severity. This assessment method can quickly screen out fluctuations that may threaten power grid stability, and the obtained severity assessment data lays the foundation for subsequent anomaly identification.
[0038] When identifying anomalies in peak load increases using severity assessment data, time series analysis tools can be used for localization. For example, if a load peak is observed to increase from the normal 300 kW to 900 kW within a certain local area over a specific time period, time series analysis tools can initially pinpoint the anomaly to a specific hourly segment, helping to further narrow down the investigation scope.
[0039] Precise labeling can be achieved by combining spatial distribution information and time series data to determine the specific time node of the peak increase. Suppose that the analysis finds that the peak increase occurs between 14:00 and 15:00 on a certain day, and is concentrated in the area near a certain substation, the specific time node can be labeled as 14:30.
[0040] When generating comprehensive analysis results for load fluctuation anomalies, precise time-point labeled data and spatial distribution information can be integrated to ultimately pinpoint the specific location of the peak increase as a particular line within a substation and at a specific time. This comprehensive analysis provides power grid maintenance personnel with clear anomaly location information, facilitating rapid response and ensuring the stability of power grid operation.
[0041] Step S103 includes: obtaining historical load records from a pre-established load data repository, organizing load change data within a specified time range, and determining daily peak points and corresponding time nodes; calculating peak increments within adjacent time periods based on the daily peak points and time nodes, analyzing peak change trends, and obtaining the continuity characteristics of peak changes; identifying abnormal locations and specific time nodes of peak increments based on the continuity characteristics of peak changes, recording the distribution of abnormal increments, and determining potential risk points; if the peak increment exceeds a preset threshold, marking the abnormal location and time node as a high-risk area and triggering a preliminary warning signal; for the high-risk area, combining historical load data and voltage instability records, using a support vector machine model for classification prediction to determine the probability of voltage instability; and generating a corresponding risk assessment level based on the probability of voltage instability.
[0042] When retrieving historical load records from a pre-established load data repository, assuming the past 30 days are selected, and daily load variation curves are compiled, it is found that daily peak points are mostly concentrated between 2 PM and 4 PM, with load values typically around 5000 MW. This data extraction provides a foundation for subsequent analysis.
[0043] The calculation of peak increments within adjacent time periods is as follows. Suppose that on a certain day the peak increases from 5000 MW to 5500 MW, an increment of 500 MW. Combining the increment data from three consecutive days, the analysis shows that the peak change trend exhibits a continuous upward trend. This trend analysis helps identify potential destabilizing factors.
[0044] When identifying abnormal locations and time points of peak increases, if the peak increase exceeds a preset threshold, such as 600 MW, the abnormal location and time point are marked as a high-risk area and a preliminary warning signal is triggered. For example, if an increase reaches 850 MW, the system will automatically mark the industrial substation as a high-risk area and send a warning signal to the dispatch center so that timely countermeasures can be taken.
[0045] When analyzing high-risk areas by combining historical load data and voltage instability records, a support vector machine (SVM) model can be used to classify and predict the probability of voltage instability. Assuming that the area has experienced multiple voltage fluctuations in the past, and considering the current load increase, the model predicts a 75% probability of voltage instability. This prediction method provides a scientific basis for risk assessment.
[0046] The aforementioned steps, from data acquisition to risk assessment, form a complete logical chain. Through the processing of historical data, analysis of peak trends, identification of anomalies, triggering of early warning signals, and the assistance of model predictions, a comprehensive understanding of load changes can be achieved, providing strong support for the stable operation of the power grid. In particular, with the accurate marking of high-risk areas and the reasonable classification of early warning levels, dispatchers can allocate resources more efficiently and reduce the possibility of system failures.
[0047] Step S104 includes: obtaining voltage instability data for local areas from the risk assessment results; using automated tools to perform hierarchical processing on the voltage instability data to obtain a set of dynamic change parameters within the area; using a convolutional neural network model to predict short-term load change trends based on the set of dynamic change parameters to determine the load fluctuation range for each area; for the load fluctuation range, if the fluctuation of a certain area exceeds a preset threshold, the area is marked as a high-risk object, and the energy regulation demand sequence of the high-risk object is obtained; using the energy regulation demand sequence and combined with the area division data, the load transferability between each area is determined to obtain a preliminary priority order list; based on the preliminary priority order list, the matching degree between the dynamic change parameters of each area and the load prediction results is obtained to determine the final energy regulation priority sequence.
[0048] Voltage fluctuation data for a specific neighborhood within a city's power grid is extracted from historical load records. Assuming that the neighborhood's voltage frequently fell below standard values during peak hours each day over the past week, automated tools are used to stratify the data by time period and fluctuation amplitude, resulting in a set of dynamically changing parameters. These parameters may include the hourly voltage drop percentage and duration; for example, a 5% voltage drop lasting 2 hours from 6:00 PM to 8:00 PM on a certain day. This stratified processing helps to clearly understand the patterns and severity of voltage instability.
[0049] When predicting short-term load change trends for dynamically changing parameter sets, convolutional neural network models can be used to analyze the periodic characteristics of cell load. For example, if the prediction shows that the peak load of a cell may increase by 10% within the next three days, with fluctuations exceeding a preset threshold of 8%, it is marked as a high-risk area. This prediction method can identify potential problem areas in advance, providing data support for subsequent adjustments.
[0050] When obtaining the energy regulation demand sequence for high-risk areas, the additional electrical resources required for the area can be calculated based on the predicted load increase. For example, if the demand sequence shows a daily increase of 500 kWh of electricity, combining this with regional data and analyzing the load margin in surrounding areas reveals 300 kWh of transferable electricity daily in neighboring areas, thus providing a preliminary assessment of the feasibility of load shifting. This analysis helps optimize resource allocation.
[0051] The initial priority list can be adjusted based on the matching degree between load forecasts and dynamic parameters for each region. For example, if a region experiences significant load fluctuations and is projected to see substantial growth, while neighboring regions show smaller fluctuations, then energy regulation should be prioritized for that region. This approach ensures that resources are directed to the areas most in need.
[0052] Step S105 includes: obtaining priority data from a pre-established energy regulation database, performing preliminary screening of high-priority areas, and determining a target area list; obtaining available energy and energy reserve data for each area based on the target area list, and determining whether the energy reserves of each area meet a preset threshold through data comparison; if they are lower than the preset threshold, they are marked as shortage areas; obtaining actual demand data for the shortage areas, and using statistical tools to calculate the difference between the actual demand and the energy reserves to obtain the energy gap value for each area; sorting the energy gaps of the high-priority areas based on the energy gap value and the priority data using a linear regression model to determine the emergency allocation order; obtaining available energy data for surrounding areas based on the emergency allocation order; if the available energy in the surrounding areas is higher than a preset threshold, generating a cross-regional energy allocation instruction to obtain the allocated energy distribution result; calculating the load balance status of each area after allocation based on the energy allocation result, and determining whether the load has reached a balanced state through data comparison to obtain the final adjustment plan.
[0053] For the step of acquiring data on available energy and energy reserves in each region and determining whether they meet a preset threshold, we can assume the preset threshold is 5,000 kWh of reserves per day. In a target region, the current reserve is only 3,000 kWh, below the threshold, and therefore it is marked as a shortage area. This comparison method intuitively reflects the degree of energy shortage in a region, providing a basis for subsequent allocation.
[0054] When calculating the difference between actual demand and energy reserves in a region with energy shortages, assuming that the actual demand in a region is 8,000 kWh while the reserves are only 3,000 kWh, the difference is 5,000 kWh, which is the energy gap value. This value quantifies the size of the gap and helps to accurately identify the regions most in need of support.
[0055] To determine the order of emergency allocation by ranking energy gaps in high-priority regions using a linear regression model, it is conceivable to input the gap values of multiple high-priority regions into the model, combine them with priority data, and then deduce that the region with the largest gap is ranked first.
[0056] When generating cross-regional energy allocation instructions, if there is an area with an energy reserve of 7,000 kWh near the area with a shortage, which is 5,000 kWh higher than the preset threshold, then 2,000 kWh can be allocated to support the area with a shortage. This allocation instruction is based on data comparison to ensure the efficiency and feasibility of energy flow.
[0057] Regarding the calculation of the load balance in each region after the load adjustment, we can assume that the load in the shortage area drops from 8000 kWh to 6000 kWh, approaching a balanced state. Meanwhile, the load in the support area remains within a safe range, indicating that the load adjustment is reasonable. This method of judgment helps verify the effectiveness of the adjustment plan.
[0058] During the final adjustment plan formulation process, load data from various regions can be repeatedly compared to ensure that all areas with shortages are covered. For example, if a region still has a small deficit after the adjustment, further adjustments will be made from other regions with sufficient reserves until the overall balance is achieved. This meticulous adjustment process reflects the comprehensiveness and practicality of the plan, and helps improve the stability and resource utilization efficiency of the power grid system.
[0059] Step S106 includes: acquiring system load balancing data, analyzing the resource allocation status in the load balancing data, determining the available capacity of backup power and energy storage devices, and generating an initial allocation scheme; calculating the coverage range of the fluctuation response based on the initial allocation scheme; if the coverage range is lower than a preset threshold, triggering a residual insufficiency determination to obtain the gap data of the fluctuation response; adjusting the resource path using an iterative optimization method for the gap data, reallocating the output ratio of the backup power and the energy storage devices, and judging the optimized resource configuration result; extracting key indicators of power supply reliability from the optimized resource configuration result; if the key indicators do not meet the standards, adjusting the resource path again to determine a new allocation combination; analyzing the matching degree between power supply reliability and the improved configuration based on the new allocation combination, and obtaining the applicability evaluation value of the final scheme by comparing historical data; and locking the resource path and allocation ratio based on the applicability evaluation value to generate complete data for the power supply reliability improvement configuration.
[0060] When calculating the coverage of the fluctuation response, assume the system needs to handle a 10% load fluctuation, or a fluctuation demand of 50 MW, while the current backup and energy storage coverage capacity is only 40 MW, below the preset threshold of 60 MW. This triggers a deficiency determination, resulting in a shortfall of 20 MW. To address this shortfall, resource paths can be adjusted through iterative optimization methods. For example, prioritize calling up backup power from nearby areas, adjust the output ratio, increase the release ratio of energy storage devices from 20 MW to 25 MW, and simultaneously reduce the output pressure on backup power.
[0061] After optimizing resource allocation, when extracting key indicators of power supply reliability, attention can be paid to the frequency and duration of power outages. For example, if the optimized outage frequency decreases from twice per month to once per month, but still falls short of the standard of 0.5 times per month, then resource paths need to be adjusted again, and new allocation combinations should be tried, such as increasing the proportion of nighttime charging for energy storage devices to ensure more available capacity during peak periods.
[0062] When analyzing the match between power supply reliability and the upgraded configuration, historical data can be compared. For example, if the previous solution had an outage frequency of 3 times per month under similar loads, while the new solution reduces it to 1 time per month, the applicability assessment value is high, indicating the solution's feasibility. Finally, resource paths are determined based on the assessment values. For instance, if the fixed energy storage device outputs an average of 25 MW per day, and the backup power supply is 15 MW, complete data is generated. This path adjustment can effectively improve power supply stability during peak periods.
[0063] In practical implementation, iterative optimization of the fluctuation response gap can start with the core solution: prioritizing the adjustment of local resource ratios. If this is still insufficient, it can be expanded to cross-regional resource allocation, forming a multi-layered guarantee. This approach ensures both efficient utilization of local resources and enhances overall reliability through external support, jointly supporting the goal of stable system operation.
[0064] In adjusting the output ratio of backup power and energy storage devices, the ratio can be dynamically allocated based on the characteristics of the load curve. For example, during off-peak hours at night, the output of backup power can be reduced while the charging capacity of energy storage devices can be increased to reserve more energy for peak hours during the day. This flexible configuration can significantly improve resource utilization efficiency and reduce waste.
[0065] Continuous optimization of power supply reliability indicators can be achieved by regularly updating the historical data comparison database to ensure the accuracy of the assessed values. This approach helps to identify potential risks in a timely manner, adjust configurations in advance, and ensure long-term operational stability.
[0066] Step S107 includes: obtaining an optimized configuration scheme that conforms to the current power grid state from a pre-established power supply reliability improvement scheme database, and determining the appropriate execution parameters; sending the optimized configuration scheme to the main control system of the control center through a real-time transmission channel, and recording the timestamp and data integrity status during the transmission process; driving the execution device to deploy the scheme for the optimized configuration scheme received by the control center, and triggering a backup scheme switch if an abnormal signal is detected during the deployment process, obtaining the deployment completion status information; starting the voltage signal acquisition module according to the deployment completion status information, continuously monitoring the voltage signal data output by the execution device, and obtaining real-time signal fluctuation records; using a support vector machine algorithm to classify the signal fluctuation records, determining whether there are abnormal fluctuations exceeding a preset threshold, and determining the signal quality assessment result; and analyzing the optimization performance achievement by combining the signal quality assessment result with historical feedback data to obtain performance verification data.
[0067] When retrieving optimized configuration schemes from a pre-established database of power supply reliability improvement schemes, the system can analyze the current power grid load distribution and historical operating data to select the most suitable scheme. Assuming the current power grid is under peak load, and the database stores multiple configuration schemes for similar scenarios, the system will prioritize schemes that have performed stably during past peak periods and extract corresponding execution parameters, such as setting the backup power supply activation ratio to 30% and adjusting the energy storage device's discharge rate to 50 kWh per hour. This approach allows for rapid adaptation to power grid conditions, ensuring the feasibility of the scheme.
[0068] When sending the plan to the control center via a real-time transmission channel, multiple verification mechanisms can be set up to ensure data integrity. Assuming the transmission is timestampd at 14:30:25 and the data integrity status shows 100%, this indicates no data loss. If data integrity is found to be lower than expected, the system will automatically retransmit, ensuring the control center receives the accurate plan.
[0069] To handle abnormal signals during deployment, a pre-set backup switch mechanism can be used. For example, if the actuator detects a voltage drop during deployment, the system will immediately switch to the backup mechanism, such as temporarily increasing the energy storage device's output to 60% to stabilize the power supply. This rapid response effectively prevents deployment interruptions.
[0070] When the voltage signal acquisition module is activated for real-time monitoring, voltage data can be collected every 5 seconds, recording the fluctuation range. Assuming the normal voltage range is 220 volts ± 5%, if 3 out of 10 consecutive data acquisitions exceed this range, the system will mark it as an abnormal fluctuation. This continuous monitoring helps to promptly identify potential problems.
[0071] When using the Support Vector Machine (SVM) algorithm to classify signal fluctuation records, the fluctuation data can be categorized into three types: normal, slightly abnormal, and severely abnormal. For example, if a signal fluctuation record shows the voltage repeatedly jumping to 230 volts, the system would classify it as slightly abnormal and generate a corresponding quality assessment report. This classification process provides data support for subsequent optimization.
[0072] When analyzing performance optimization, historical feedback data can be used to assess whether the current solution meets expectations. For example, if historical data shows a voltage stability rate of 95% for similar solutions, while the current solution only achieves 90%, the system will record this difference and generate performance verification data. This comparative analysis helps identify areas for improvement.
[0073] When updating database parameter configurations, the system can adjust the solution parameters based on performance verification data. For example, if verification data indicates that the energy storage device's output ratio is too low, the system will increase it to 65% and simultaneously store the latest execution performance records. This dynamic update mechanism continuously optimizes the database content, improving the adaptability of future solutions.
Claims
1. A method for optimizing voltage stability in a distribution network, characterized in that, Includes the following steps: S101 collects electricity consumption data and voltage signals in local areas of the power distribution network in real time through a sensor network, processes these data to identify instantaneous fluctuation characteristics, and obtains instantaneous load fluctuation sequences; S102 uses a support vector machine algorithm to classify the fluctuation type and severity based on the obtained instantaneous load fluctuation sequence, and determines the potential location and time point of peak increase; S103 obtains the determined peak increase location and time point, integrates historical load data for trend analysis, and determines that if the peak increase exceeds the preset threshold, it triggers a voltage instability warning and obtains a voltage instability risk assessment result. Based on the obtained voltage instability risk assessment results, S104 extracts the dynamic change parameters of the local area, predicts short-term load demand through a convolutional neural network model, and determines the priority sequence of energy regulation. S105 selects a high-priority area from the determined energy regulation priority sequence, calculates the difference between available energy reserves and demand, and obtains a load balance adjustment scheme; S106 allocates backup power and energy storage equipment resources according to the obtained load balancing adjustment scheme. If the remaining fluctuation response is insufficient after allocation, the resource path is iteratively optimized to determine the final power supply reliability improvement configuration. S107 obtains the final power supply reliability improvement configuration and transmits it to the actuator in the control center in real time. It monitors the voltage signal feedback after execution and obtains optimized performance verification data.
2. The method for predicting power consumption and optimizing charging of multiple unmanned aerial vehicles as described in claim 1, characterized in that, Step S101 includes: The power consumption data and voltage signals of local areas of the power distribution network are collected in real time through sensor networks to obtain raw data streams; After preprocessing the original data stream, a purified data signal is obtained, and time series segments related to instantaneous fluctuations in the purified data signal are extracted to determine the time point and amplitude range of the fluctuations.
3. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S102 includes: Instantaneous load fluctuation sequence data is obtained from the monitoring system, and noise and outliers are removed through preprocessing to obtain cleaned load sequence data; For the cleaned load sequence data, the support vector machine algorithm is used to classify the fluctuation type, determine whether the fluctuation is periodic or sudden, and obtain the classification result of the fluctuation type. Based on the classified fluctuation type results, the severity of the fluctuation type is analyzed. If the fluctuation amplitude exceeds the preset threshold, it is marked as high severity, and severity assessment data is obtained. Using the severity assessment data, anomalies with increased peak values are identified, and time series analysis tools are used to locate the potential locations of the increased peak values to obtain the spatial distribution information of the anomalies. Based on the spatial distribution information of the anomalies and combined with time series data, the specific time nodes of the peak increase are determined, and the precise labeled data of the time nodes are obtained. By using the precisely labeled data of the time nodes and the spatial distribution information, a comprehensive analysis result of load fluctuation anomalies is generated to determine the final peak increase location and time point.
4. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S103 includes: Historical load records are retrieved from a pre-established load data repository, load change data within a specified time range is compiled, and daily peak points and corresponding time nodes are determined. Based on the daily peak points and time nodes, calculate the peak increment within adjacent time periods, analyze the peak change trend, and obtain the continuity characteristics of peak change; Based on the continuous characteristics of the peak value changes, identify the abnormal locations and specific time points of the peak value increments, record the distribution of the abnormal increments, and determine potential risk points. If the peak increment exceeds a preset threshold, the abnormal location and time point will be marked as a high-risk area, triggering a preliminary warning signal. For the high-risk areas, a support vector machine model is used to classify and predict the probability of voltage instability by combining historical load data and voltage instability records. The corresponding risk assessment level is generated based on the probability of voltage instability occurring.
5. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S104 includes: Voltage instability data for local areas is obtained from the risk assessment results. Automated tools are used to perform layered processing on the voltage instability data to obtain a set of dynamically changing parameters within the area. Based on the set of dynamically changing parameters, a convolutional neural network model is used to predict short-term load change trends and determine the load fluctuation range for each region. For the load fluctuation range, if the fluctuation of a certain area exceeds a preset threshold, the area is marked as a high-risk object, and the energy regulation demand sequence of the high-risk object is obtained; By combining the energy regulation demand sequence with regional division data, the load transferability between regions is determined, and a preliminary priority order list is obtained. Based on the preliminary priority list, the matching degree between the dynamic change parameters of each region and the load forecast results is obtained, and the final energy regulation priority sequence is determined.
6. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S105 includes: Priority data is obtained from a pre-established energy regulation database, and high-priority areas are initially screened to determine a list of target areas; Based on the target area list, obtain the available energy and energy reserve data for each area, and determine whether the energy reserves of each area meet the preset threshold by comparing the data. If they are lower than the preset threshold, they are marked as scarce areas. For the aforementioned areas with severe energy shortages, actual demand data is obtained, and statistical tools are used to calculate the difference between the actual demand and the energy reserves to obtain the energy gap value for each area. Based on the energy gap value and the priority data, the energy gap of the high-priority areas is sorted using a linear regression model to determine the emergency allocation order. For the emergency allocation order, obtain the available energy data of the surrounding area. If the available energy in the surrounding area is higher than a preset threshold, generate a cross-regional energy allocation instruction and obtain the allocated energy distribution result. Based on the energy allocation results, the load balance status of each region after the adjustment is calculated, and the load balance status is determined by data comparison to obtain the final adjustment plan.
7. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S106 includes: Acquire system load balance data, analyze the resource allocation status in the load balance data, determine the available capacity of backup power and energy storage devices, and generate an initial allocation scheme; Calculate the coverage area of the fluctuation response based on the initial allocation scheme; If the coverage area is lower than a preset threshold, a residual insufficiency determination is triggered to obtain gap data of the fluctuation response; For the aforementioned gap data, an iterative optimization method is used to adjust the resource path, reallocate the output ratio of the backup power supply and the energy storage device, and determine the optimized resource allocation result. Extract key indicators of power supply reliability from the optimized resource allocation results; If the aforementioned key indicators are not met, the resource path will be adjusted again to determine a new allocation combination; Based on the new allocation combination, the matching degree between power supply reliability and the improved configuration is analyzed, and the applicability evaluation value of the final solution is obtained by comparing historical data. Based on the applicability assessment value, resource paths and allocation ratios are locked, and complete data for power supply reliability improvement configuration is generated.
8. The distribution network voltage stability optimization method as described in claim 1, characterized in that, Step S107 includes: Obtain optimized configuration schemes that conform to the current power grid status from a pre-established database of power supply reliability improvement schemes, and determine the appropriate execution parameters; The optimized configuration scheme is sent to the main control system of the control center through a real-time transmission channel, and the timestamp and data integrity status are recorded during the transmission process. Based on the optimized configuration scheme received by the control center, the drive execution device deploys the scheme. If an abnormal signal is detected during the deployment process, the backup scheme is switched and the deployment completion status information is obtained. Based on the deployment completion status information, the voltage signal acquisition module is activated to continuously monitor the voltage signal data output by the execution device and obtain real-time signal fluctuation records. The signal fluctuation records are classified using a support vector machine algorithm to determine whether there are abnormal fluctuations exceeding a preset threshold, and to determine the signal quality assessment result. By combining the signal quality assessment results with historical feedback data, the performance optimization results are analyzed to obtain performance verification data.