Hierarchical and hierarchical autonomous balancing method for active power distribution network based on multiple scenarized targets

Abstract

The invention discloses an active power distribution network hierarchical autonomous balancing method based on scenarized multiple targets, and belongs to the technical field of power system power distribution network operation control. Comprising the following steps: defining a balance range according to a unit, grid and partition three-level architecture, and dividing a receiving type group, a balance type group and a sending type group by double indexes; a typical scene is described in multiple dimensions, an extreme load day is selected through a convex hull algorithm, a conventional day is selected in normal distribution, and new energy output is fitted in combination with Beta distribution to generate a credible curve; a differential active balance strategy is adopted, candidate measures are matched based on scene features, an optimal scheme is selected by adopting cost-benefit scores, and hierarchical power flow optimization cross-voltage hierarchy collaboration is realized; and full-dimensionally evaluating adequacy and safety, and dynamically iteratively adjusting the full-dimensionally evaluated adequacy and safety. The method improves the power grid balance capability and the new energy consumption rate, reduces the operation cost, guarantees the system stability, and is suitable for the intelligent operation of the active power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power system distribution network operation control technology, and in particular to a hierarchical and autonomous balancing method for active distribution networks based on scenario-based multi-objectives. Background Technology

[0002] With the rapid development of new energy power generation technologies, the penetration rate of distributed power sources (such as photovoltaic and wind power) in active distribution networks is constantly increasing. However, their intermittent and random characteristics have led to a significant increase in uncertainty on both the source and load sides. The traditional distribution network's reliance on passive dispatch for balancing is no longer sufficient to meet the demand for new energy consumption. At the same time, the resource endowments of different voltage levels of the distribution network (10kV unit layer, 110kV grid layer, 220kV zone layer) are significantly different. The local area has insufficient regulation capacity, cross-regional power exchange is difficult, and the linkage mechanism between overall planning and local response is lacking, which further exacerbates the difficulty of balancing supply and demand. In existing technologies, the balancing strategies of active distribution networks often lack scenario-based multi-objective design. They fail to develop differentiated measures for different scenarios such as peak output and power balance, and data preprocessing methods are inefficient (e.g., outlier removal and standardization methods are simplistic). Group segmentation does not fully integrate key indicators such as source-load density and generation-to-consumption ratio. Evaluation systems focus only on a single dimension (e.g., safety or economy), lacking a comprehensive evaluation across sufficiency, safety, cleanliness, and economy. Furthermore, they lack a dynamic closed-loop optimization mechanism, making it difficult to adapt to annual changes in grid topology and source-load characteristics. To address these issues, a scenario-based multi-objective hierarchical autonomous balancing method for active distribution networks is needed. This method, through systematic data preprocessing, scientific group segmentation, precise scenario characterization, differentiated active balancing strategies, and comprehensive state evaluation, achieves coordinated balancing across the entire network and local areas, improving renewable energy absorption capacity and grid operating efficiency. This has become an urgent need in the current active distribution network field.

[0003] Patent CN119674933A discloses a grid-based generation-grid-load-storage hierarchical and partitioned balancing model, including: S1, dividing the power distribution system into three levels: power consumption grid, power supply unit, and power supply grid; S2, determining the grid interface parameters of the partition boundary of the power consumption grid and the related power flow constraint equations; S3, determining the grid interface parameters of the partition boundary of the power supply unit and the related power flow constraint equations; S4, determining the grid interface parameters of the partition boundary of the power supply grid and the related power flow constraint equations. This invention can effectively control the maximum load level that the transmission system needs to supply, distinguish the differences in power supply reliability requirements between conventional loads and controllable loads, take effective control measures, reduce the maximum load demand of the system, and improve the load characteristics of the system. It plays an important role in improving the utilization rate of power grid assets and equipment and the efficiency of asset investment. It can effectively promote the planning and realization of the integrated generation-grid-load-storage system in the power distribution system. The above-mentioned technical solutions focus on dividing the power consumption grid, power supply unit and other levels and determining the interface parameters and constraint equations. They lack the characterization of typical scenarios such as peak output and power balance, and cannot effectively match the different balance requirements caused by the intermittency and randomness of new energy sources. They have not yet broken through the limitations of traditional passive dispatch. Therefore, it is imperative for those skilled in the art to solve the aforementioned technical problems. Summary of the Invention

[0004] To this end, the present invention provides a scenario-based multi-objective hierarchical autonomous balancing method for active distribution networks. Through scenario-based multi-objective design and hierarchical autonomous balancing, it effectively solves the supply and demand imbalance caused by the intermittency of new energy sources, improves the grid regulation capacity and new energy absorption rate, takes into account safety, economy and cleanliness, dynamically iterates to adapt to changes in source, grid and load, and optimizes grid operation efficiency.

[0005] To achieve the above objectives, the present invention provides a hierarchical and autonomous balancing method for active distribution networks based on scenario-based multi-objectives, characterized by the following steps: Step S1. Basic data preparation;

[0006] Step S2. Hierarchical and balanced group division;

[0007] Step S3. Depicting typical balance scenarios in multiple dimensions;

[0008] Step S4. Execute the differentiated proactive balancing strategy;

[0009] Step S5. Evaluation of the balance state across all dimensions.

[0010] By adopting the above technical solutions, a complete technical framework for a scenario-based, multi-objective, hierarchical autonomous balancing method for active distribution networks was constructed, covering five key steps: basic data preparation, hierarchical group division, multi-dimensional scenario characterization, differentiated strategy execution, and comprehensive evaluation. This framework overcomes the limitations of existing distribution network balancing methods, which are fragmented and lack systematicity. Through a hierarchical architecture design, it achieves precise control of resources at different voltage levels; combined with scenario-based multi-objective design, it considers various operational needs such as peak output and power balance; and the closed-loop design ensures the method's dynamic adaptability. Compared to traditional passive dispatching methods, this framework can more effectively address the uncertainties on both the source and load sides brought about by the high penetration of renewable energy, improving the overall balancing capacity and operational efficiency of the power grid. It has significant innovation and practicality, providing a systematic solution for intelligent balancing of active distribution networks.

[0011] Furthermore, in step S1, source load basic data, historical load data, new energy output data, power grid topology data, and load forecast data of the target area are collected;

[0012] The data is preprocessed and standardized, outliers are removed and missing data is added, and key indicators are calculated, specifically as follows:

[0013] Step S11. Abnormal data removal:

[0014] Calculate the mean of the data series Standard deviation Remove The outliers, among which, It is the arithmetic mean of the data sequence. It is the total number of data points. It is the first The values ​​of the original data points, It is the sample standard deviation of the data sequence;

[0015] Step S12. Gradient mutation detection:

[0016] Calculate gradient Load data retrieval Data on new energy output Remove | |> The mutation value, where, It is the first Time's up Gradient change over time, It is the first Data values ​​at any given time It is the first Data values ​​at any given time It is the first The timestamp of the moment It is the timestamp of time i. It is the gradient mutation threshold;

[0017] Step S13. Missing data completion:

[0018] linear interpolation ;

[0019] 2nd-3rd order polynomial fitting Solve for the coefficients using the least squares method, and substitute them into the time t_k to obtain the supplementary value. ;

[0020] Eliminate dimensional differences , The mean, Let be the standard deviation. After standardization, the mean is 0 and the variance is 1. It is missing. Time data values, yes Known data values ​​at time [time] yes Known data values ​​at time [time] yes The timestamp of the moment yes The timestamp of the moment It is missing. Time stamp;

[0021] Step S14. Calculation of Key Indicators

[0022] Load density ,in, S represents the total load capacity, and S represents the area of ​​the region.

[0023] Power density ,in, This represents the total installed capacity of new energy sources;

[0024] usage ratio ,in, This represents the total power generation from new energy sources. This represents the total electricity consumption of the region.

[0025] By adopting the above technical solutions, the basic data preparation stage is refined. Through scientific data preprocessing and key indicator calculation, accurate and reliable data support is provided for subsequent steps. Outlier removal employs a combination of the 3σ criterion and gradient mutation detection, handling outliers in normally distributed data while addressing abrupt changes in non-stationary time-series data. Missing data supplementation uses linear interpolation or polynomial fitting based on the duration of missing data to ensure data continuity. Z-score standardization eliminates dimensional differences and improves data comparability. Key indicator calculations (load density, power density, generation-to-utilization ratio) provide a quantitative basis for balance type classification, avoiding the subjectivity of qualitative classification in existing technologies. This step significantly improves data quality, ensuring the accuracy of subsequent group classification and scenario characterization, and resolving the problem of balance strategy failure caused by coarse data preprocessing in existing technologies.

[0026] Furthermore, in step S2, the balance range is defined according to the three-level architecture of cell, mesh, and partition;

[0027] The balance type is classified using a dual-index determination method;

[0028] Divide the balance groups based on load characteristics, balance type, and energy flow direction, and output a list;

[0029] The balancing range is defined from bottom to top using a three-level architecture: units corresponding to the 10kV voltage level, grids corresponding to the 110kV voltage level, and zones corresponding to the 220kV voltage level. The balancing type is divided into three types: receiving type, balancing type, and transmitting type, based on the ratio of load density to power density or the generation-consumption ratio.

[0030] Balance groups are divided based on source load endowment and energy flow direction;

[0031] When the ratio of load density to power density is >1.2, it is a receiving type; when the ratio is 0.8-1.2, it is a balanced type; and when the ratio is <0.8, it is a transmitting type.

[0032] When the generation-to-utilization ratio is <0.8, it is a receiving type; when the generation-to-utilization ratio is 0.8-1.2, it is a balanced type; and when the generation-to-utilization ratio is >1.2, it is a sending type.

[0033] By adopting the above technical solution, a three-tier architecture—units, grids, and zones—corresponding to 10kV, 110kV, and 220kV voltage levels, is implemented to achieve hierarchical management of resources, meeting the voltage-level management requirements of actual distribution network operation. A dual-index judgment method, using the load density to power density ratio and generation-consumption ratio, classifies balance types, avoiding the bias of single indicators and making the classification of receiving, balancing, and transmitting types more accurate. Combining source-load endowment with energy flow direction to group resources allows for targeted matching of regional characteristics, providing a foundation for subsequent differentiated strategies. This method solves the problem of low balancing efficiency caused by unreasonable regional division in existing technologies, achieving precise resource allocation, improving cross-regional power mutual assistance capabilities, and providing a scientific basis for the hierarchical management of active distribution networks.

[0034] Further, in step S3:

[0035] Step S31. Select typical days through cluster analysis to obtain the load characteristic curves of typical days;

[0036] Step S32. Determine the confidence interval of new energy output through probability analysis and obtain the typical daily power output characteristic curve;

[0037] Step S33. Combine the load characteristic curve and the output characteristic curve to generate a typical sun-duck curve and clarify the equilibrium boundary;

[0038] The clustering analysis in step S31 uses the convex hull algorithm, specifically:

[0039] Suppose the annual load curve dataset is as follows:

[0040] ;in This represents the load data at the mth hour on the i-th day, where t is the time and P is the load power.

[0041] The convex hull CH(X) of dataset X is calculated as follows:

[0042] ,in, Here, k is the convex combination coefficient, and k is the number of vertices of the convex hull. Three types of extreme typical days are selected: the day with the maximum annual load, the day with the minimum annual load, and the day with the maximum load peak-to-valley difference.

[0043] Based on the probability distribution characteristics of the daily average load, dates that conform to the normal distribution confidence interval are selected as typical days for regular operation scenarios. The specific calculation method is as follows:

[0044] Calculate the daily average load in the annual load data Construct daily average load sequence ;

[0045] Calculate the mean of the sequence. with standard deviation :

[0046] ;

[0047] ;

[0048] Select confidence interval The dates within this range are used as typical days and satisfy the following conditions:

[0049] ;

[0050] Two categories of typical days are selected: typical weekdays and typical restdays. The load characteristic curves for these typical days are then output. It is the first The average daily load for the day, where m is the number of whole hours in a day;

[0051] In step S32, the confidence level of new energy output is determined according to the balance scenario: peak output balance is 0.8-0.95, carrying capacity balance is 0.1-0.2, and power balance is 0.5-0.6.

[0052] By adopting the above technical solution, typical days are selected through cluster analysis (extreme days for convex hull algorithm and regular days for normal distribution) to cover both extreme and daily operation scenarios of the distribution network. Probabilistic analysis is used to determine the confidence interval for renewable energy output, and the output characteristics are fitted using a Beta distribution model, fully considering the intermittency and randomness of renewable energy. Duck curves for typical days are generated to clarify the equilibrium boundary, providing a clear basis for subsequent strategy execution. This step solves the problem of incomplete scenario coverage or inaccurate characterization in existing technologies, making the scenario model closer to actual operating conditions, ensuring that differentiated strategies can accurately match the needs of different scenarios, and improving renewable energy absorption capacity and grid operation stability.

[0053] Furthermore, in step S33, a probability model for new energy output is established, the confidence level is determined according to different balance scenarios, and a typical daily reliable output curve is generated;

[0054] The probability model for new energy output includes:

[0055] The new energy output probability model adopts the Beta distribution model, which is suitable for fitting the output of distributed power sources such as photovoltaic and wind power, which have intermittent and random characteristics. The specific structure and mathematical calculation method are as follows:

[0056] The new energy output probability model takes historical output data as input, fits the output probability density function through a Beta distribution, and outputs a typical daily reliable output curve by combining the confidence requirements of different equilibrium scenarios. The specific calculation method is as follows:

[0057]

[0058] In the formula, for Typical daily performance is reliable and reliable; For new energy installed capacity; Output rate at time t (Values ​​range [0,1]) Let be the probability density function of the Beta distribution;

[0059] Next, the historical power output data is normalized, and the power output rate at each time point is calculated. ;

[0060] Beta distribution parameter estimation: The method of moments is used to calculate the shape parameters α and β of the Beta distribution.

[0061] ;

[0062] ;

[0063] In the formula, For output rate The mean, For output rate The variance;

[0064] Quantities of output rate are calculated based on the confidence requirements of different equilibrium scenarios. ,satisfy:

[0065]

[0066] In the formula, C represents the confidence level. γ(t) is the quantile of the output rate at time t;

[0067] Calculate the reliable output at each moment based on the installed capacity. Plot a typical daily reliable output curve, and combine the typical daily load characteristic curve with the reliable output curve of new energy sources to plot a typical scenario duck curve and clarify the balance boundary.

[0068] By adopting the above technical solution, candidate measures are screened through scenario matching rules (based on feature vectors of adjustment cycle, frequency, and cost), and autonomous selection is achieved using a cost-benefit scoring method, ensuring the economy and effectiveness of the measures. Voltage level coordinated measures adopt energy storage, demand response, and load coupling methods to address different cycle demands, and cross-level optimization uses a hierarchical power flow model with the goal of minimizing network losses. This strategy overcomes the limitations of existing technologies that rely on single balancing measures and lack intelligent selection, achieving precise matching between measures and scenarios, improving the efficiency of cross-regional resource allocation, reducing operating costs, and solving the problem of difficult distribution network balancing under high penetration of new energy sources.

[0069] Furthermore, in step S4, the opposite side is adjusted with a margin for infrequent adjustment by adjusting the operating mode;

[0070] Frequent adjustments to achieve load complementarity between adjacent areas are achieved through flexible interconnection;

[0071] By optimizing the power supply range, the long-term, high-frequency balance requirements can be met.

[0072] The scenario matching rules and interval coordination measure selection algorithm for the differentiated active balancing strategy adopt a cost-benefit priority algorithm to achieve autonomous judgment, specifically:

[0073] Define scene feature vector Among them, the adjustment cycle T for the operation mode adjustment is greater than 24 hours, the adjustment frequency F does not exceed 1 time / week, and the cost threshold C does not exceed 0.1 yuan / kWh;

[0074] The adjustment cycle T of flexible interconnection does not exceed 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold is between 0.1 yuan / kWh and 0.5 yuan / kWh;

[0075] The adjustment cycle T for optimizing the power supply range is greater than 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold C does not exceed 0.3 yuan / kWh;

[0076] The cost-benefit scoring method is used to achieve autonomous selection of flexible interconnection, optimized power supply range, and differentiated active balancing. The specific calculation method is as follows:

[0077] ;

[0078] in, The measures are scored, and the higher the score, the higher the priority for selection. This represents the increase in balancing capacity after the implementation of the measures. To balance overall demand, To measure investment costs, For investment budget, These are the weighting coefficients;

[0079] The voltage level coordination measures in the differentiated active balancing strategy include:

[0080] For short-cycle adjustment needs, chemical energy storage charging and discharging or demand response are initiated; for scenarios with complementary load characteristics within the region, load coupling optimization is adopted.

[0081] Optimize cross-voltage level resource allocation by using a power boost output method;

[0082] Improve the local power grid regulation capability by adopting a distributed power supply approach;

[0083] Voltage level coordination is implemented using a hierarchical power flow optimization algorithm, which optimizes cross-level power flow by combining the resource endowments of different voltage levels. Specifically:

[0084] A hierarchical power flow optimization model is constructed with minimizing network loss as the objective function:

[0085] ;

[0086] Constraints: Power balance constraints: Voltage constraint: Line capacity constraints: ;

[0087] in, The total network loss is [amount]. For the number of voltage levels, The number of lines at each level, For the first Level 1 Line current, For line resistance, To provide power, For load power, For the first Hierarchical network loss, For the first The power exchange between hierarchical levels and between higher and lower levels; Node voltage; This is the maximum allowable current for the line.

[0088] By adopting the above technical solution, indicators are selected from four dimensions: sufficiency, safety, cleanliness, and economy. Differentiated evaluation methods are used based on the type of grid balance and the needs of the scenario (e.g., for receiving grids, sufficiency is the primary focus, while for transmitting grids, both sufficiency and cleanliness are considered), and a weighted comprehensive score is used for balanced grids. This system addresses the problem of existing technologies having only one evaluation dimension, comprehensively reflecting the grid balance status and providing a scientific basis for subsequent adjustments. Objective evaluation is achieved through quantitative indicators (capacity-to-load ratio, renewable energy penetration rate, loss load ratio, etc.), ensuring the reliability of the evaluation results and assisting in grid operation optimization and decision-making.

[0089] Furthermore, in step S5, the comprehensive balance state evaluation selects evaluation indicators from four dimensions: sufficiency, safety, cleanliness, and economy, specifically as follows:

[0090] Differentiated indicators are used for evaluation based on the balance type and scenario requirements. For receiving types, adequacy indicators are used; for sending types, adequacy and cleanliness indicators are used; and for balanced types, adequacy, safety, cleanliness, and economy are comprehensively evaluated.

[0091] For scenarios with differentiated demand, the focus is on evaluating safety indicators during peak power output balancing, while for scenarios with balanced power output, the focus is on evaluating economic indicators.

[0092] The adequacy is characterized by the capacity-to-load ratio, which represents the degree of matching between the power grid's supply capacity and load demand. The specific calculation method is as follows:

[0093] ;

[0094] The capacity-to-generation ratio characterizes the degree of matching between the grid's transmission capacity and the installed capacity of new energy sources. The calculation method is as follows:

[0095] ;

[0096] In the formula, k is the new energy output coefficient, which is determined according to the scenario confidence level;

[0097] The cleanliness is characterized by the renewable energy penetration rate, which represents the proportion of renewable energy power generation in total electricity consumption. The specific calculation method is as follows:

[0098] ;

[0099] The aforementioned safety index characterizes the power supply reliability of the power grid under fault or overload conditions by the proportion of lost load. The specific calculation method is as follows:

[0100] ;

[0101] The economic indicators characterize the utilization efficiency of power grid equipment through the average load rate of the equipment, and the specific calculation method is as follows:

[0102] ;

[0103] The balanced group comprehensive evaluation calculates a comprehensive evaluation score by weighting and summing four categories of indicators: sufficiency, safety, cleanliness, and economy. The specific calculation method is as follows:

[0104] ;

[0105] in, The standardized scores are for the four categories of indicators; These are the weighting coefficients.

[0106] By adopting the above technical solutions, the construction scale of each voltage level is clearly defined through overall balancing, local bottlenecks are resolved through regional scenario-based balancing, and millisecond-level data interaction is achieved through 5G collaborative communication, establishing a "overall coordination - regional response" linkage mechanism. This implementation process ensures the operability and coordination of the strategy, and the point-to-area linkage between the overall and regional levels addresses the lack of coordination in existing technologies. 5G communication ensures efficient information transmission, and hierarchical power flow optimization achieves optimal resource allocation across levels. This step improves the efficiency of strategy execution, ensures the rapid response capability of the power grid in different scenarios, and provides technical support for the intelligent operation of active distribution networks.

[0107] Furthermore, the implementation steps of the differentiated proactive balancing strategy include:

[0108] The global power balance adopts a hierarchical power flow optimization algorithm, which combines the resource endowment of the 10kV unit layer, the 110kV grid layer and the 220kV partition layer to optimize the cross-level power flow direction and clarify the annual construction scale of each voltage level.

[0109] Regional scenario-based balancing addresses local bottleneck issues by filtering candidate measures through scenario matching rules and employing a cost-benefit priority algorithm to select the optimal solution and formulate specific optimization strategies.

[0110] The 5G collaborative communication channel utilizes its millisecond-level low latency to enable real-time interaction of source and load data and balancing commands between the overall coordination center and local area nodes, ensuring the efficiency of information transmission.

[0111] Establish a “global coordination-regional response” linkage mechanism, which coordinates cross-regional resource allocation at the global level and enables regions to respond quickly based on global instructions or their own scenario needs, through a closed-loop process of “intra-level optimization-inter-level power interaction-global optimal verification”.

[0112] By adopting the above technical solution, adjustments are triggered when evaluation indicators fail to reach preset thresholds (such as excessively high loss load ratio or excessively low self-balancing coefficient). Adjustment methods include group division optimization, scenario characterization updates, and replacement of balancing measures. Annual updated data covers changes in load forecasting, renewable energy installations, and grid topology. This mechanism addresses the lack of dynamic adaptability in existing technologies, ensuring that the method can be adjusted promptly according to changes in grid and source-load characteristics, maintaining the timeliness and effectiveness of the strategy, and achieving continuous optimization of active distribution network balancing strategies.

[0113] Furthermore, adjustments are triggered when the evaluation indicators do not reach the preset thresholds, i.e., when the proportion of lost load is >0.5%, balancing measures are adjusted, and when the self-balancing coefficient is <0.8, the group division is adjusted.

[0114] The specific adjustment methods are as follows: when adjusting the group division, the source-load density ratio is recalculated and adjacent receiving and sending groups are merged to optimize the boundary; when adjusting the scenario characterization, the historical photovoltaic output data of the past year is supplemented to regenerate the typical daily curve; when adjusting the balancing measures, high-cost chemical energy storage is replaced with demand response and load coupling.

[0115] The annually updated data includes load forecast data, newly installed capacity of new energy sources, power grid topology changes, and equipment aging parameters.

[0116] By adopting the above technical solution, and through a process of repeated group division, scenario characterization, active balancing, and state evaluation, the balancing strategy and power grid construction scale are dynamically adjusted. This closed-loop optimization mechanism addresses the lack of long-term optimization in existing technologies, ensuring that the method can continuously adapt to annual changes in the power grid and maintain its technological advancement and practicality. Through annual iterations, the power grid balancing capacity is continuously improved, resource allocation is optimized, operating costs are reduced, and a guarantee is provided for the long-term stable operation of the active distribution network.

[0117] Furthermore, the closed-loop optimization cycle is to carry out a full-process iterative optimization once a year, that is, to repeat the process of group division, scenario characterization, active balancing, and state evaluation in order to dynamically adjust the balancing strategy and the scale of power grid construction.

[0118] The beneficial effects of this invention are as follows:

[0119] 1. This invention constructs a complete technical framework for the hierarchical and autonomous balancing of active distribution networks based on scenario-based multi-objectives. It covers five key steps: basic data preparation, hierarchical group division, multi-dimensional scenario characterization, differentiated strategy execution, and full-dimensional evaluation. It breaks through the limitations of the fragmented and unsystematic distribution network balancing methods in the prior art, achieves precise control of resources at different voltage levels, takes into account multiple operational needs, ensures the dynamic adaptability of the method, and more effectively responds to the uncertainties on both the source and load sides brought about by the high penetration of new energy, thereby improving the overall balancing capability and operational efficiency of the power grid.

[0120] 2. This invention refines the basic data preparation process by employing a combination of the 3σ criterion and gradient mutation detection for outlier data removal, hierarchical missing data supplementation, Z-score standardization, and key indicator calculation. This provides accurate and reliable data support for subsequent steps, avoids the subjectivity of qualitative classification in existing technologies, significantly improves data quality, ensures the accuracy of subsequent group division and scenario characterization, and solves the problem of balancing strategy failure caused by coarse data preprocessing in existing technologies.

[0121] 3. This invention defines the balance range by adopting a three-level architecture of units, grids, and partitions corresponding to different voltage levels, and classifies balance types by combining a dual-index judgment method. It also classifies balance groups based on the characteristics of specific regions, thereby solving the problem of low balance efficiency caused by unreasonable regional division in the prior art. This enables precise allocation of resources, improves cross-regional power mutual assistance capabilities, and provides a scientific basis for the hierarchical management of active distribution networks.

[0122] 4. This invention selects typical days through cluster analysis, determines the confidence interval of new energy output through probability analysis, and generates duck curves for typical scenarios. Combined with scenario matching rules and cost-benefit scoring methods, it realizes the autonomous selection of differentiated active balancing strategies, matches the needs of different scenarios, improves the new energy absorption capacity and grid operation stability, breaks through the limitations of existing technologies that have single balancing measures and lack intelligent selection, reduces operating costs, and solves the problem of difficult distribution network balancing under high new energy penetration. Attached Figure Description

[0123] Figure 1 This is a schematic diagram of the equilibrium system of the present invention;

[0124] Figure 2 This is a schematic diagram illustrating the active balancing of the active distribution network based on the source-grid-load-storage coordinated optimization of the present invention.

[0125] Figure 3 This is a schematic diagram of the active balancing architecture of the present invention. Detailed Implementation

[0126] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0127] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0128] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0129] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0130] Please see Figure 1 , Figure 2 and Figure 3 As shown, this invention provides a hierarchical and graded autonomous balancing method for active distribution networks based on scenario-based multi-objectives, including:

[0131] Step S1. Basic data preparation;

[0132] Step S2. Hierarchical and balanced group division;

[0133] Step S3. Depicting typical balance scenarios in multiple dimensions;

[0134] Step S4. Execute the differentiated proactive balancing strategy;

[0135] Step S5. Evaluation of the balance state across all dimensions.

[0136] A hierarchical and graded autonomous balancing method for active distribution networks based on multiple objectives across all scenarios is established. This method improves four optimization methods for balancing status evaluation standards by dividing balancing groups into hierarchical levels, characterizing typical balancing scenarios, and implementing site-specific proactive balancing strategies and measures. This addresses the problem of new sources and loads having high uncertainty and difficulty participating in balancing, achieving supply and demand coordination between zones and voltage levels, reducing the scale of grid infrastructure expansion, and improving the economical and efficient operation of the grid.

[0137] Based on zoning, grids, and units, the energy flow direction of the planning objects is determined and balance groups of each voltage level are constructed according to the results of new source load survey and prediction. This provides a basis for subsequent formulation of active balancing strategies, differentiated planning schemes, and targeted evaluations.

[0138] By analyzing the annual load curve, representative typical days are selected to clarify the load curve of typical days; by analyzing the characteristics and probability of random power output through historical data, the power output confidence interval under different balance scenarios is clarified; by combining the typical daily load curve and the power output confidence interval, a duck curve for typical scenarios is drawn to clarify the balance boundary.

[0139] By coordinating various flexible resources and optimizing the grid supply load curve through multi-regional autonomy, inter-regional mutual assistance, and coordination among different voltage levels, the grid can operate more efficiently and reduce the increase in grid construction, operation, and control costs caused by the access of new energy sources.

[0140] The evaluation indicators for the balance state are improved by taking into account different needs such as power supply, reliability, new energy consumption, and equipment operation efficiency improvement, and differentiated evaluation indicators are set according to different construction scenarios.

[0141] In step S1, source load basic data, historical load data, new energy output data, power grid topology data and load forecast data of the target area are collected;

[0142] The data is preprocessed and standardized, outliers are removed and missing data is added, and key indicators are calculated, specifically as follows:

[0143] Step S11. Abnormal data removal:

[0144] Calculate the mean of the data series Standard deviation Remove The outliers, among which, It is the arithmetic mean of the data sequence. It is the total number of data points. It is the first The values ​​of the original data points, It is the sample standard deviation of the data sequence;

[0145] Step S12. Gradient mutation detection:

[0146] Calculate gradient Load data retrieval Data on new energy output Remove | |> The mutation value, where, It is the first Time's up Gradient change over time, It is the first Data values ​​at any given time It is the first Data values ​​at any given time It is the first The timestamp of the moment It is the timestamp of time i. It is the gradient mutation threshold;

[0147] Step S13. Missing data completion:

[0148] linear interpolation ;

[0149] 2nd-3rd order polynomial fitting Solve for the coefficients using the least squares method, and substitute them into the time t_k to obtain the supplementary value. ;

[0150] Eliminate dimensional differences , The mean, Let be the standard deviation. After standardization, the mean is 0 and the variance is 1. It is missing. Time data values, yes Known data values ​​at time [time] yes Known data values ​​at time [time] yes The timestamp of the moment yes The timestamp of the moment It is missing. Time stamp;

[0151] Step S14. Calculation of Key Indicators

[0152] Load density ,in, S represents the total load capacity, and S represents the area of ​​the region.

[0153] Power density ,in, This represents the total installed capacity of new energy sources;

[0154] usage ratio ,in, This represents the total power generation from new energy sources. This represents the total electricity consumption of the region.

[0155] In step S2, the balance range is defined by a three-level architecture of cells, meshes, and partitions;

[0156] The balance type is classified using a dual-index determination method;

[0157] Divide the balance groups based on load characteristics, balance type, and energy flow direction, and output a list;

[0158] The balancing range is defined from bottom to top using a three-level architecture: units corresponding to the 10kV voltage level, grids corresponding to the 110kV voltage level, and zones corresponding to the 220kV voltage level. The balancing type is divided into three types: receiving type, balancing type, and transmitting type, based on the ratio of load density to power density or the generation-consumption ratio.

[0159] Balance groups are divided based on source load endowment and energy flow direction;

[0160] When the ratio of load density to power density is >1.2, it is a receiving type; when the ratio is 0.8-1.2, it is a balanced type; and when the ratio is <0.8, it is a transmitting type.

[0161] When the generation-to-utilization ratio is <0.8, it is a receiving type; when the generation-to-utilization ratio is 0.8-1.2, it is a balanced type; and when the generation-to-utilization ratio is >1.2, it is a sending type.

[0162] In step S3, step S31: Select typical days through cluster analysis to obtain the load characteristic curve of typical days;

[0163] Step S32. Determine the confidence interval of new energy output through probability analysis and obtain the typical daily power output characteristic curve;

[0164] Step S33. Combine the load characteristic curve and the output characteristic curve to generate a typical sun-duck curve and clarify the equilibrium boundary;

[0165] The clustering analysis in step S31 uses the convex hull algorithm, specifically:

[0166] Suppose the annual load curve dataset is as follows:

[0167] ;in This represents the load data at the mth hour on the i-th day, where t is the time and P is the load power.

[0168] The convex hull CH(X) of dataset X is calculated as follows:

[0169] ,in, Here, k is the convex combination coefficient, and k is the number of vertices of the convex hull. Three types of extreme typical days are selected: the day with the maximum annual load, the day with the minimum annual load, and the day with the maximum load peak-to-valley difference.

[0170] Based on the probability distribution characteristics of the daily average load, dates that conform to the normal distribution confidence interval are selected as typical days for regular operation scenarios. The specific calculation method is as follows:

[0171] Calculate the daily average load in the annual load data Construct daily average load sequence ;

[0172] Calculate the mean of the sequence. with standard deviation :

[0173] ;

[0174] ;

[0175] Select confidence interval The dates within this range are used as typical days and satisfy the following conditions:

[0176] ;

[0177] Two categories of typical days are selected: typical weekdays and typical restdays. The load characteristic curves for these typical days are then output. It is the first The average daily load for the day, where m is the number of whole hours in a day;

[0178] The confidence level of new energy output in step S32 is determined according to the balance scenario: peak output balance is 0.8-0.95, carrying capacity balance is 0.1-0.2, and power balance is 0.5-0.6.

[0179] In step S33, a probability model for new energy output is established, the confidence level is determined according to different balance scenarios, and a typical daily reliable output curve is generated.

[0180] The probability model for new energy output includes:

[0181] The new energy output probability model adopts the Beta distribution model, which is suitable for fitting the output of distributed power sources such as photovoltaic and wind power, which have intermittent and random characteristics. The specific structure and mathematical calculation method are as follows:

[0182] The new energy output probability model takes historical output data as input, fits the output probability density function through a Beta distribution, and outputs a typical daily reliable output curve by combining the confidence requirements of different equilibrium scenarios. The specific calculation method is as follows:

[0183]

[0184] In the formula, for Typical daily performance is reliable and reliable; For new energy installed capacity; Output rate at time t (Values ​​range [0,1]) Let be the probability density function of the Beta distribution;

[0185] Next, the historical power output data is normalized, and the power output rate at each time point is calculated. ;

[0186] Beta distribution parameter estimation: The method of moments is used to calculate the shape parameters α and β of the Beta distribution.

[0187] ;

[0188] ;

[0189] In the formula, For output rate The mean, For output rate The variance;

[0190] Quantities of output rate are calculated based on the confidence requirements of different equilibrium scenarios. ,satisfy:

[0191]

[0192] In the formula, C represents the confidence level. γ(t) is the quantile of the output rate at time t;

[0193] Calculate the reliable output at each moment based on the installed capacity. Plot a typical daily reliable output curve, and combine the typical daily load characteristic curve with the reliable output curve of new energy sources to plot a typical scenario duck curve and clarify the balance boundary.

[0194] In step S4, the opposite side is adjusted with a margin for infrequent adjustments by adjusting the operating mode;

[0195] Frequent adjustments to achieve load complementarity between adjacent areas are achieved through flexible interconnection;

[0196] By optimizing the power supply range, the long-term, high-frequency balance requirements can be met.

[0197] The scenario matching rules and interval coordination measure selection algorithm for the differentiated active balancing strategy adopt a cost-benefit priority algorithm to achieve autonomous judgment, specifically:

[0198] Define scene feature vector Among them, the adjustment cycle T for the operation mode adjustment is greater than 24 hours, the adjustment frequency F does not exceed 1 time / week, and the cost threshold C does not exceed 0.1 yuan / kWh;

[0199] The adjustment cycle T of flexible interconnection does not exceed 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold is between 0.1 yuan / kWh and 0.5 yuan / kWh;

[0200] The adjustment cycle T for optimizing the power supply range is greater than 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold C does not exceed 0.3 yuan / kWh;

[0201] The cost-benefit scoring method is used to achieve autonomous selection of flexible interconnection, optimized power supply range, and differentiated active balancing. The specific calculation method is as follows:

[0202] ;

[0203] in, The measures are scored, and the higher the score, the higher the priority for selection. This represents the increase in balancing capacity after the implementation of the measures. To balance overall demand, To measure investment costs, For investment budget, These are the weighting coefficients;

[0204] The voltage level coordination measures in the differentiated active balancing strategy include:

[0205] For short-cycle adjustment needs, chemical energy storage charging and discharging or demand response are initiated; for scenarios with complementary load characteristics within the region, load coupling optimization is adopted.

[0206] Optimize cross-voltage level resource allocation by using a power boost output method;

[0207] Improve the local power grid regulation capability by adopting a distributed power supply approach;

[0208] Voltage level coordination is implemented using a hierarchical power flow optimization algorithm, which optimizes cross-level power flow by combining the resource endowments of different voltage levels. Specifically:

[0209] A hierarchical power flow optimization model is constructed with minimizing network loss as the objective function:

[0210] ;

[0211] Constraints: Power balance constraints: Voltage constraint: Line capacity constraints: ;

[0212] in, The total network loss is [amount]. For the number of voltage levels, The number of lines at each level, For the first Level 1 Line current, For line resistance, To provide power, For load power, For the first Hierarchical network loss, For the first The power exchange between hierarchical levels and between higher and lower levels; Node voltage; This is the maximum allowable current for the line.

[0213] In step S5, the comprehensive balance state evaluation selects evaluation indicators from four dimensions: sufficiency, safety, cleanliness, and economy, specifically as follows:

[0214] Differentiated indicators are used for evaluation based on the balance type and scenario requirements. For receiving types, adequacy indicators are used; for sending types, adequacy and cleanliness indicators are used; and for balanced types, adequacy, safety, cleanliness, and economy are comprehensively evaluated.

[0215] For scenarios with differentiated demand, the focus is on evaluating safety indicators during peak power output balancing, while for scenarios with balanced power output, the focus is on evaluating economic indicators.

[0216] The adequacy is characterized by the capacity-to-load ratio, which represents the degree of matching between the power grid's supply capacity and load demand. The specific calculation method is as follows:

[0217] ;

[0218] The capacity-to-generation ratio characterizes the degree of matching between the grid's transmission capacity and the installed capacity of new energy sources. The calculation method is as follows:

[0219] ;

[0220] In the formula, k is the new energy output coefficient, which is determined according to the scenario confidence level;

[0221] The cleanliness is characterized by the renewable energy penetration rate, which represents the proportion of renewable energy power generation in total electricity consumption. The specific calculation method is as follows:

[0222] ;

[0223] The aforementioned safety index characterizes the power supply reliability of the power grid under fault or overload conditions by the proportion of lost load. The specific calculation method is as follows:

[0224] ;

[0225] The economic indicators characterize the utilization efficiency of power grid equipment through the average load rate of the equipment, and the specific calculation method is as follows:

[0226] ;

[0227] The balanced group comprehensive evaluation calculates a comprehensive evaluation score by weighting and summing four categories of indicators: sufficiency, safety, cleanliness, and economy. The specific calculation method is as follows:

[0228] ;

[0229] in, The standardized scores are for the four categories of indicators; These are the weighting coefficients.

[0230] The implementation steps of the differentiated active balancing strategy include:

[0231] The global power balance adopts a hierarchical power flow optimization algorithm, which combines the resource endowment of the 10kV unit layer, the 110kV grid layer and the 220kV partition layer to optimize the cross-level power flow direction and clarify the annual construction scale of each voltage level.

[0232] Regional scenario-based balancing addresses local bottleneck issues by filtering candidate measures through scenario matching rules and employing a cost-benefit priority algorithm to select the optimal solution and formulate specific optimization strategies.

[0233] The 5G collaborative communication channel utilizes its millisecond-level low latency to enable real-time interaction of source and load data and balancing commands between the overall coordination center and local area nodes, ensuring the efficiency of information transmission.

[0234] Establish a “global coordination-regional response” linkage mechanism, which coordinates cross-regional resource allocation at the global level and enables regions to respond quickly based on global instructions or their own scenario needs, through a closed-loop process of “intra-level optimization-inter-level power interaction-global optimal verification”.

[0235] Adjustments are triggered when evaluation indicators fail to reach preset thresholds, i.e., when the loss load ratio is >0.5%, balancing measures are adjusted, and when the self-balancing coefficient is <0.8, group division is adjusted.

[0236] The specific adjustment methods are as follows: when adjusting the group division, the source-load density ratio is recalculated and adjacent receiving and sending groups are merged to optimize the boundary; when adjusting the scenario characterization, the historical photovoltaic output data of the past year is supplemented to regenerate the typical daily curve; when adjusting the balancing measures, high-cost chemical energy storage is replaced with demand response and load coupling.

[0237] The annually updated data includes load forecast data, newly installed capacity of new energy sources, power grid topology changes, and equipment aging parameters.

[0238] The closed-loop optimization cycle is to carry out a full-process iterative optimization once a year, which means repeating the process of group division, scenario characterization, active balancing, and state evaluation to dynamically adjust the balancing strategy and the scale of power grid construction.

[0239] In one embodiment, refer to Figure 2 and Figure 3 It can be seen that the active balancing of the active distribution network in the source-grid-load-storage collaborative optimization is divided into three parts: hierarchical, hierarchical balancing group division, balancing scenario characterization, and active balancing.

[0240] 1. Hierarchical and tiered balanced group division

[0241] (1) Division method

[0242] 1) Determine the type of balance

[0243] Based on the results of new energy source and load surveys and load forecasts, energy source and load endowments are estimated from bottom to top according to units (corresponding to 10 kV voltage level), grids (corresponding to 110 kV voltage level), and zones (corresponding to 220 kV voltage level), and the energy flow type of units, grids, and zones is clarified.

[0244] There are three types of power balancing: transmitting, balancing, and receiving. There are two methods for classifying these types: One is based on the load-to-power density ratio: a ratio greater than 1.2 indicates receiving, between 1.2 and 0.8 indicates balancing, and less than 0.8 indicates transmitting. The other method is based on the generation-to-consumption ratio: a ratio less than 0.8 indicates receiving, between 0.2 and 0.8 indicates balancing, and greater than 1.2 indicates transmitting. Method one applies to power balancing, while method two applies to electricity balancing.

[0245] Table 1. Method for classifying balance types

[0246]

[0247] 2) Divide into balanced groups

[0248] Based on the differences in load characteristics in different regions and in combination with the balancing type, balancing groups are divided into three levels: unit, grid, and zone.

[0249] (2) The significance of balanced group partitioning

[0250] First, by classifying the balancing types into hierarchical and tiered categories, the differences in source and load endowments, flexible resource conditions, and load characteristics between different voltage levels and regions are clearly displayed, providing a basis for coordinating resource conditions at different levels, combining short-term and long-term perspectives, and optimizing the formulation of tiered balancing schemes.

[0251] Secondly, it facilitates differentiated balance assessments. Energy receiving types are mainly assessed based on capacity ratio, while energy sending and energy balance types require a comprehensive assessment of both capacity ratio and capacity generation ratio, thereby improving the rationality of balance calculations.

[0252] Table 2 Recommended range of capacity ratio

[0253]

[0254] Table 3 Recommended range for capacity ratio

[0255]

[0256] Third, it provides a basis for subsequent differentiated planning schemes. For energy transmission-oriented scenarios, the focus should be on the access and transmission of clean energy. Based on the forecast of the exploitable capacity of clean energy and its output characteristics, the target scale and capacity of distribution network construction should be calculated, and the transmission capacity of distribution facilities should be verified. For energy receiving-oriented scenarios, the focus should be on the aggregation and interaction of diverse loads and the activation of dormant grid resources, improving the utilization efficiency of grid equipment, enhancing the grid's self-healing capabilities, and building a robust and reliable target grid structure. For energy balancing-oriented scenarios, the focus should be on improving the grid's flexible adjustment capabilities and the coordinated control capabilities of power generation, grid, load, and storage. Through digital empowerment and technological innovation, the efficiency of energy resource utilization should be improved, and the economic benefits of energy storage configuration should be enhanced while ensuring reliable power supply.

[0257] In one embodiment, the purpose of characterizing a typical balance scenario is to provide boundaries for power supply and demand balance analysis and calculation. The process and method are as follows:

[0258] Step 1: Obtain the annual load curve of the balancing object, and select typical days through cluster analysis to obtain the load characteristic curve determined for each typical day. Step 2: Analyze the annual output curves of different types of renewable energy sources, and determine the confidence level of renewable energy output corresponding to each typical day through probability analysis to obtain a reliable output characteristic curve for each typical day. Step 3: Combine the determined load characteristic curve and the reliable output characteristic curve to generate a duck-shaped curve for each typical day. Step 4: Compare the duck-shaped curves for each typical day, reduce some scenarios, and obtain the balance boundary.

[0259] Typical day selection

[0260] Selecting appropriate representative days is crucial for accurately defining the equilibrium boundary. Selecting too many representative days increases workload and reduces efficiency. Selecting too few or too many representative days will result in unrepresentative data and negatively impact the equilibrium results. To quickly, reasonably, and effectively obtain representative days, and to avoid mechanically applying seasonal patterns, peak-valley patterns, and relying on manual experience, data analysis methods such as the convex hull algorithm and normal distribution are introduced.

[0261] Selection of Confidence Level for Photovoltaic Output

[0262] Photovoltaic power generation is characterized by intermittency and randomness. The output power is affected by factors such as geographical location, air quality, climate conditions, and equipment aging. The output results obtained by conventional methods are difficult to be universally applicable. In order to objectively cover the output of distributed generation and take into account the over-allocation effect of extreme weather on distributed generation capacity, a probabilistic model of photovoltaic power output at the hour is studied based on the definition of confidence level.

[0263] During the balancing process, the selection of photovoltaic output confidence level should comprehensively consider the relationship between reliability, economy and estimation accuracy based on the needs of different balancing scenarios.

[0264] Table 4 Recommended Confidence Values ​​of Photovoltaic Output in Typical Balanced Scenarios

[0265]

[0266] 3. Active balancing

[0267] Based on the characteristics of the duck curve and considering various resources, differentiated strategies are formulated to optimize the duck curve through proactive participation of distributed power sources, proactive adjustment of operating modes, proactive response of diverse loads, and proactive collaborative optimization across different voltage levels. The balancing results are evaluated and iterated to ultimately determine the scale for each side: power source, grid, load, and storage. Corresponding policy mechanisms are proposed based on the problems identified during the balancing process.

[0268] (1) Active balancing process

[0269] Based on typical daily load characteristic curves, renewable energy output characteristic curves, horizontal annual load forecasts, and renewable energy scale calculations, the grid-supply load participating in the balancing process is determined. Differentiated balancing strategies are developed for different scenarios, optimizing load curves through measures such as regional autonomy, inter-regional complementarity, and coordination across different voltage levels. The balancing results are evaluated and iterated to ultimately obtain the planning results for power sources, grid, load, and energy storage. Furthermore, corresponding policy mechanisms are proposed based on problems encountered during the balancing process.

[0270] (2) Scenarios for applying proactive balancing measures

[0271] Different balancing methods are applicable to different scenarios and have different balancing costs. In the process of balancing electricity supply and demand, the balancing objectives of different scenarios should be fully considered. Based on the load characteristics and frequency of occurrence of the scenario, the balancing strategy should be optimized according to local conditions, and the balancing method should be selected scientifically and rationally.

[0272] Table 5 Applicable Scenarios for Different Balancing Methods

[0273]

[0274] (3) Evaluation indicators for different balance scenarios

[0275] Active power distribution networks face dynamic, diverse, and differentiated operating scenarios. Traditional deterministic evaluation indicators may not fully reflect the system's balance status and are insufficient to coordinate multiple objectives such as sufficiency and economy. Therefore, based on the construction objectives of active power distribution networks, and building upon traditional power supply and demand balance evaluation indicators, a comprehensive evaluation of the balance status is conducted from three dimensions: safety margin, renewable energy consumption, and economic operation, in order to reflect the effectiveness of proactive balancing.

[0276] Table 6 Recommended Evaluation Indicators for Equilibrium State

[0277]

[0278] (III) Balanced Content

[0279] 1. Balanced Object

[0280] Three balancing levels: The balancing is carried out at three levels: unit level, grid level, and zone level, corresponding to 10 kV balancing, 110 kV balancing, and 220 kV balancing, respectively.

[0281] Three energy flow types: For each level, energy flow is divided, representing the resource differences and source-load relationships within the same level.

[0282] Two types of balancing are carried out: load power balancing and power supply balancing are carried out at each level.

[0283] The active balancing of the active distribution network with coordinated optimization of source, grid, load and storage includes global balancing and regional scenario-based balancing.

[0284] The target of the overall balance is the entire city of Yixing. The purpose of the overall balance is to clarify the annual scale of each voltage level of the distribution network based on the distribution of power sources and the scale of development, reserve a reasonable safety margin, and guide the construction of the power grid to be moderately advanced. The overall balance includes the 110 kV power grid balance and the 10 (20) kV power grid balance.

[0285] Regional scenario-based balancing combines diagnostic analysis results with targeted balancing of local power grid problems, such as difficulties in integrating and absorbing new energy sources in local areas, power grid bottlenecks due to rapid load growth, and low equipment utilization efficiency, and proposes local power grid optimization solutions.

[0286] In one embodiment, the present invention is based on the concept of "3×3×2 balanced system" and "source-grid-load-storage coordinated optimization". The implementation process is divided into three major stages: basic data preparation, hierarchical core implementation, and balance evaluation and iterative optimization. The steps of each stage are clear and closely connected, as detailed below:

[0287] Phase 1: Basic Data Preparation (Prerequisite for Implementation)

[0288] Full-dimensional data collection: Collect basic source and load data for the target area (such as the entire region and each zone, grid, and unit), including: new source and load survey results (distributed photovoltaic, energy storage, residential / industrial / commercial load distribution and capacity), historical load data (annual load curves for the past 3-5 years), new energy output data (historical output records of photovoltaic / wind power), power grid topology data (connection methods and equipment parameters of 220kV / 110kV / 10kV voltage levels), and load forecast data (horizontal annual load growth rate and load density distribution).

[0289] Data preprocessing and standardization: Remove abnormal data (such as sudden changes in output / load caused by equipment failure), and use interpolation to supplement missing data; standardize data of different formats (such as unifying the unit of load density to MW / km² and the unit of power density to MW / km²); calculate key indicators (generation-to-consumption ratio, load density to power density ratio) to provide data support for subsequent group division and scenario characterization.

[0290] Phase Two: Layered and Hierarchical Core Implementation (Core Processes)

[0291] Step 1: Hierarchical and balanced group division

[0292] Define the balance levels and scope: Based on the three-level architecture of "unit (10kV) - grid (110kV) - zone (220kV)," define the balance scope of each level from bottom to top to ensure coverage of the entire power grid and local areas.

[0293] Determining the equilibrium type: The "dual-index judgment method" is used to classify the equilibrium type.

[0294] Power balance scenario: determined by the ratio of load density to power supply density (>1.2 for receiving type, 0.8-1.2 for balanced type, <0.8 for transmitting type).

[0295] Power balance scenario: determined by "generation-to-consumption ratio" (<0.8 is receiving type, 0.8-1.2 is balanced type, >1.2 is sending type).

[0296] Divide the load balancing groups: Based on the differences in load characteristics, balancing types and energy flow at each level, divide the areas with similar source load endowments within the same voltage level into independent balancing groups, forming a division result of "three-level architecture + three types + multiple groups", and output the "Balanced Group Division List" (marking the voltage level, balancing type, coverage area and source load capacity of each group).

[0297] Step 2: Depicting Typical Balanced Scenarios in Multiple Dimensions

[0298] Typical day selection: Based on the annual load curve, cluster analysis is performed using the convex hull algorithm (extreme operation mode) and the normal distribution method (economic operation mode) to select representative typical days covering weekdays, rest days, holidays and extreme load periods, and generate typical day load characteristic curves for each balance group.

[0299] Determining the confidence interval for new energy output: Analyzing the randomness and volatility of historical photovoltaic / wind power output data, establishing an output probability model through probabilistic statistical methods, determining the confidence level according to different balance scenarios (0.8-0.95 for peak output balance, 0.1-0.2 for carrying capacity balance, and 0.5-0.6 for power balance), and generating a reliable output curve for typical days.

[0300] Balance boundary and duck curve plotting: Combining typical daily load characteristic curves and reliable new energy output curves, duck curves for typical scenarios of each group are plotted to clarify the peak / valley time of output, the peak-valley difference of load and the supply-demand balance boundary, and output the "Typical Balance Scenario Description".

[0301] Step 3: Execution of the differentiated proactive balancing strategy

[0302] The objects and contents of the balance are defined as follows: "Load power balance" and "Power supply power balance" are carried out respectively for the three levels of "unit-grid-zone". The balance covers the balance of the whole region (clarifying the construction scale of each voltage level year by year and guiding the power grid to be moderately ahead) and the balance of regional scenarios (focusing on problems such as local heavy load and difficulty in the consumption of new energy).

[0303] Balancing Measures Selection and Implementation: Based on the balancing type, scenario characteristics, and cost optimization principle for each group, select appropriate balancing measures.

[0304] 1) Regional autonomy: For short-cycle adjustment needs (such as surplus photovoltaic output at midday), activate chemical energy storage charging and discharging or demand response (low-frequency adjustment scenarios); for scenarios with complementary load characteristics within the region, adopt load coupling optimization;

[0305] 2) Inter-regional coordination: When there is adjustment margin on the other side, load transfer is achieved through operation mode adjustment; when the loads of adjacent areas are complementary and frequent adjustments are required, flexible interconnection technology is activated; for long-term, high-frequency balancing needs, an optimized power supply range scheme is adopted;

[0306] Voltage level coordination: To meet long-term balance requirements, power supply boosting or distributed access methods are adopted to optimize resource allocation across voltage levels.

[0307] "Point-to-surface" coordinated execution: In overall balance, the construction scale of each level is coordinated and a safety margin is reserved; in regional scenario-based balance, special optimization solutions (such as load cutover and distributed power aggregation and control) are formulated for local bottleneck problems (such as low equipment utilization and difficulty in new energy consumption), and the coordinated response of the whole area and local areas is realized through 5G collaborative communication channels.

[0308] Step 4: Evaluation of the overall equilibrium state

[0309] Evaluation indicators are selected based on four dimensions: adequacy (capacity-to-load ratio, capacity-to-generation ratio), safety (proportion of lost load, power restoration time), cleanliness (penetration rate of new energy, proportion of green electricity), and economy (self-balancing coefficient, average equipment load rate). Differentiated evaluation indicators are selected in combination with the type of balance and the needs of the scenario (e.g., for receiving type, the capacity-to-load ratio is the main evaluation factor, while for sending type, both the capacity-to-load ratio and capacity-to-generation ratio are evaluated simultaneously).

[0310] Evaluation of Implementation and Result Analysis: Collect operational data after the implementation of the balancing strategy, calculate the values ​​of each evaluation index, compare with the preset standard values ​​(such as the 110kV capacity-to-load ratio of 1.5-2.0 and the target value of new energy penetration rate), analyze the balancing effect (such as the increase in new energy consumption rate and the degree of optimization of equipment load rate), and identify the links that do not meet the standards (such as the proportion of loss load in local areas exceeding the standard).

[0311] Phase 3: Iterative Optimization (Closed-Loop Improvement)

[0312] Problem rectification and strategy adjustment: Optimize technical solutions based on the issues identified in the evaluation.

[0313] If the group division is unreasonable and results in low balancing efficiency, readjust the group boundaries (e.g., merge adjacent receiving and sending groups for optimization).

[0314] If the scenario is not accurately described, resulting in poor strategy adaptability, supplement historical data to regenerate typical daily curves and output confidence intervals;

[0315] If the cost of balancing measures is too high, replace them with a lower-cost solution that is more suitable for the scenario (such as replacing chemical energy storage with demand response + load coupling).

[0316] Long-term dynamic optimization: Every year, based on load growth and new energy installations, the basic data is updated, and the process of "group division - scenario characterization - active balancing - status evaluation" is repeated to dynamically adjust the balancing strategy and the scale of power grid construction, forming a closed-loop management mechanism of "implementation - evaluation - optimization".

[0317] Working principle: Reliable source-load data is obtained through abnormal data removal, missing data supplementation, standardization processing, and key indicator calculation in the basic data preparation stage. Then, the balance range is defined by a three-level architecture of unit grid partitioning. The dual-index judgment method is used to classify the balance types of receiving, sending, and balancing, and groups are formed in combination with source-load endowment. Then, cluster analysis is used to select extreme and normal typical days, and probability analysis is used to determine the confidence interval of new energy output to generate typical scenario curves and clarify the balance boundary. Then, based on the scenario feature vector matching operation mode, the flexible interconnection is adjusted to optimize the power supply range and other differentiated measures. The cost-benefit scoring method is used to select the optimal solution and cross-voltage level coordination is achieved through hierarchical power flow optimization. Finally, a full-dimensional evaluation is carried out from four dimensions: adequacy, safety, cleanliness, and economy. If the preset threshold is not reached, the group division, scenario characterization, or balancing measures are adjusted. Iterative optimization is carried out every year.

[0318] To address the uncertainty on both the source and load sides caused by the intermittent and random nature of new energy power output, a Beta distribution model is used to probabilistically fit the output of distributed power sources such as photovoltaics and wind power. The shape parameters α and β are calculated using the method of moments estimation. Combined with the confidence levels of different balance scenarios, such as taking 0.8-0.95 for peak power output balance, a reliable power output curve for typical days is generated. At the same time, the convex hull algorithm is used to screen extreme typical days such as the annual maximum load day and minimum load day, and the normal distribution method is used to screen regular typical days, covering extreme and daily operation scenarios and accurately characterizing the source and load characteristics.

[0319] To address the problem of strategy failure caused by the coarseness of traditional data preprocessing, an outlier removal algorithm combining the 3σ criterion and gradient mutation detection is adopted to handle outliers and mutations, as well as short-term linear interpolation, long-term polynomial fitting, and Z-score standardization algorithms to ensure data quality. Then, by calculating key indicators such as load density, power density, and generation-to-utilization ratio, a quantitative basis for group division is provided.

[0320] To address the issue of unreasonable regional division, a dual-index judgment method is adopted, which divides the regions into receiving, balancing, and transmitting groups based on the source-load density ratio and the generation-consumption ratio. Combined with a three-level architecture of 10kV units, 110kV grids, and 220kV zones, hierarchical resource management is achieved. When the self-balancing coefficient is <0.8, group adjustment is triggered to improve balancing efficiency.

[0321] To address the issue of uneconomical measure selection, candidate measures are matched using the scenario feature vector S = (adjustment period T, frequency F, cost C). Then, the optimal solution is selected using the cost-benefit scoring algorithm Sscore = 0.6 (balance capacity improvement / total demand - 0.4 investment cost / budget), thus solving the problems of traditional measures being too simplistic and costly.

[0322] To address the issue of poor cross-level coordination, a hierarchical power flow optimization algorithm is adopted. Based on ADMM, it optimizes power flow across voltage levels and combines it with 5G low-latency communication to achieve global coordination and real-time data interaction between regional nodes, establishing a "global coordination-regional response" linkage mechanism to form a closed-loop process of "intra-level optimization-inter-level interaction-global verification". Finally, adjustments are triggered when indicators such as the proportion of lost load > 0.5% fail to meet the standards. The strategy and construction scale are iterated and optimized annually to ensure that the method continuously adapts to changes in source and load. The combination of these algorithms and mechanisms effectively overcomes the limitations of traditional distribution network balancing methods, achieving intelligent, efficient, and economical balance of active distribution networks.

[0323] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A hierarchical and graded autonomous balancing method for active distribution networks based on scenario-based multi-objectives, characterized in that, Includes the following steps: Step S1. Basic data preparation; Step S2. Hierarchical and balanced group division; Step S3. Depicting typical balance scenarios in multiple dimensions; Step S4. Execute the differentiated proactive balancing strategy; Step S5. Evaluation of the balance state across all dimensions.

2. The hierarchical and autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 1, characterized in that, In step S1, source load basic data, historical load data, new energy output data, power grid topology data and load forecast data of the target area are collected; The data is preprocessed and standardized, outliers are removed and missing data is added, and key indicators are calculated, specifically as follows: Step S11. Abnormal data removal: Calculate the mean of the data series Standard deviation Remove The outliers, among which, It is the arithmetic mean of the data sequence. It is the total number of data points. It is the first The values ​​of the original data points, It is the sample standard deviation of the data sequence; Step S12. Gradient mutation detection: Calculate gradient Load data retrieval Data on new energy output Remove | |> The mutation value, where, It is the first Time's up Gradient change over time, It is the first Data values ​​at any given time It is the first Data values ​​at any given time It is the first The timestamp of the moment It is the timestamp of time i. It is the gradient mutation threshold; Step S13. Missing data completion: linear interpolation ; 2nd-3rd order polynomial fitting Solve for the coefficients using the least squares method, and substitute them into the time t_k to obtain the supplementary value. ; Eliminate dimensional differences , The mean, Let be the standard deviation. After standardization, the mean is 0 and the variance is 1. It is missing. Time data values, yes Known data values ​​at time [time] yes Known data values ​​at time [time] yes The timestamp of the moment yes The timestamp of the moment It is missing. Time stamp; Step S14. Calculation of Key Indicators Load density ,in, S represents the total load capacity, and S represents the area of ​​the region. Power density ,in, This represents the total installed capacity of new energy sources; usage ratio ,in, This represents the total power generation from new energy sources. This represents the total electricity consumption of the region.

3. The hierarchical and graded autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 2, characterized in that, In step S2, the balance range is defined by a three-level architecture of cells, meshes, and partitions; The balance type is classified using a dual-index determination method; Divide the balance groups based on load characteristics, balance type, and energy flow direction, and output a list; The balancing range is defined from bottom to top using a three-level architecture: units corresponding to the 10kV voltage level, grids corresponding to the 110kV voltage level, and zones corresponding to the 220kV voltage level. The balancing type is divided into three types: receiving type, balancing type, and transmitting type, based on the ratio of load density to power density or the generation-consumption ratio. Balance groups are divided based on source load endowment and energy flow direction; When the ratio of load density to power density is >1.2, it is a receiving type; when the ratio is 0.8-1.2, it is a balanced type; and when the ratio is <0.8, it is a transmitting type. When the generation-to-utilization ratio is <0.8, it is a receiving type; when the generation-to-utilization ratio is 0.8-1.2, it is a balanced type; and when the generation-to-utilization ratio is >1.2, it is a sending type.

4. The hierarchical and autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 1, characterized in that, In step S3: Step S31. Select typical days through cluster analysis to obtain the load characteristic curves of typical days; Step S32. Determine the confidence interval of new energy output through probability analysis and obtain the typical daily power output characteristic curve; Step S33. Combine the load characteristic curve and the output characteristic curve to generate a typical sun-duck curve and clarify the equilibrium boundary; The clustering analysis in step S31 uses the convex hull algorithm, specifically: Suppose the annual load curve dataset is as follows: ;in This represents the load data at the mth hour on the i-th day, where t is the time and P is the load power. The convex hull CH(X) of dataset X is calculated as follows: ,in, Here, k is the convex combination coefficient, and k is the number of vertices of the convex hull. Three types of extreme typical days are selected: the day with the maximum annual load, the day with the minimum annual load, and the day with the maximum load peak-to-valley difference. Based on the probability distribution characteristics of the daily average load, dates that conform to the normal distribution confidence interval are selected as typical days for regular operation scenarios. The specific calculation method is as follows: Calculate the daily average load in the annual load data Construct daily average load sequence ; Calculate the mean of the sequence. with standard deviation : ； ； Select confidence interval The dates within this range are used as typical days and satisfy the following conditions: ； Two categories of typical days are selected: typical weekdays and typical restdays. The load characteristic curves for these typical days are then output. It is the first The average daily load for the day, where m is the number of whole hours in a day; In step S32, the confidence level of new energy output is determined according to the balance scenario: peak output balance is 0.8-0.95, carrying capacity balance is 0.1-0.2, and power balance is 0.5-0.

6.

5. The scenario-based multi-objective active distribution network hierarchical autonomous balancing method according to claim 4, characterized in that, In step S33, a probability model for new energy output is established, the confidence level is determined according to different balance scenarios, and a typical daily reliable output curve is generated. The probability model for new energy output includes: The new energy output probability model adopts the Beta distribution model, which is suitable for fitting the output of distributed power sources such as photovoltaic and wind power, which have intermittent and random characteristics. The specific structure and mathematical calculation method are as follows: The new energy output probability model takes historical output data as input, fits the output probability density function through a Beta distribution, and outputs a typical daily reliable output curve by combining the confidence requirements of different equilibrium scenarios. The specific calculation method is as follows: In the formula, for Typical daily performance is reliable and reliable; For new energy installed capacity; Output rate at time t (Values ​​range [0,1]) Let be the probability density function of the Beta distribution; Next, the historical power output data is normalized, and the power output rate at each time point is calculated. ; Beta distribution parameter estimation: The method of moments is used to calculate the shape parameters α and β of the Beta distribution. ； ； In the formula, For output rate The mean, For output rate The variance; Quantities of output rate are calculated based on the confidence requirements of different equilibrium scenarios. ,satisfy: In the formula, C represents the confidence level. γ(t) is the quantile of the output rate at time t; Calculate the reliable output at each moment based on the installed capacity. Plot a typical daily reliable output curve, and combine the typical daily load characteristic curve with the reliable output curve of new energy sources to plot a typical scenario duck curve and clarify the balance boundary.

6. The hierarchical and graded autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 1, characterized in that, In step S4, the opposite side is adjusted with a margin for infrequent adjustments by adjusting the operating mode; Frequent adjustments to achieve load complementarity between adjacent areas are achieved through flexible interconnection; By optimizing the power supply range, the long-term, high-frequency balance requirements can be met. The scenario matching rules and interval coordination measure selection algorithm for the differentiated active balancing strategy adopt a cost-benefit priority algorithm to achieve autonomous judgment, specifically: Define scene feature vector Among them, the adjustment cycle T for the operation mode adjustment is greater than 24 hours, the adjustment frequency F does not exceed 1 time / week, and the cost threshold C does not exceed 0.1 yuan / kWh; The adjustment cycle T of flexible interconnection does not exceed 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold is between 0.1 yuan / kWh and 0.5 yuan / kWh; The adjustment cycle T for optimizing the power supply range is greater than 24 hours, the adjustment frequency F is greater than 1 time / week, and the cost threshold C does not exceed 0.3 yuan / kWh; The cost-benefit scoring method is used to achieve autonomous selection of flexible interconnection, optimized power supply range, and differentiated active balancing. The specific calculation method is as follows: ； in, The measures are scored, and the higher the score, the higher the priority for selection. This represents the increase in balancing capacity after the implementation of the measures. To balance overall demand, To measure investment costs, For investment budget, These are the weighting coefficients; The voltage level coordination measures in the differentiated active balancing strategy include: For short-cycle adjustment needs, chemical energy storage charging and discharging or demand response are initiated; for scenarios with complementary load characteristics within the region, load coupling optimization is adopted. Optimize cross-voltage level resource allocation by using a power boost output method; Improve the local power grid regulation capability by adopting a distributed power supply approach; Voltage level coordination is implemented using a hierarchical power flow optimization algorithm, which optimizes cross-level power flow by combining the resource endowments of different voltage levels. Specifically: A hierarchical power flow optimization model is constructed with minimizing network loss as the objective function: ； Constraints: Power balance constraints: Voltage constraint: Line capacity constraints: ; in, The total network loss is [amount]. For the number of voltage levels, The number of lines at each level, For the first Level 1 Line current, For line resistance, To provide power, For load power, For the first Hierarchical network loss, For the first The power exchange between hierarchical levels and between higher and lower levels; Node voltage; This is the maximum allowable current for the line.

7. The hierarchical and graded autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 3, characterized in that, In step S5, the comprehensive balance state evaluation selects evaluation indicators from four dimensions: sufficiency, safety, cleanliness, and economy, specifically as follows: Differentiated indicators are used for evaluation based on the balance type and scenario requirements. For receiving types, adequacy indicators are used; for sending types, adequacy and cleanliness indicators are used; and for balanced types, adequacy, safety, cleanliness, and economy are comprehensively evaluated. For scenarios with differentiated demand, the focus is on evaluating safety indicators during peak power output balancing, while for scenarios with balanced power output, the focus is on evaluating economic indicators. The adequacy is characterized by the capacity-to-load ratio, which represents the degree of matching between the power grid's supply capacity and load demand. The specific calculation method is as follows: ； The capacity-to-generation ratio characterizes the degree of matching between the grid's transmission capacity and the installed capacity of new energy sources. The calculation method is as follows: ； In the formula, k is the new energy output coefficient, which is determined according to the scenario confidence level; The cleanliness is characterized by the renewable energy penetration rate, which represents the proportion of renewable energy power generation in total electricity consumption. The specific calculation method is as follows: ； The aforementioned safety index characterizes the power supply reliability of the power grid under fault or overload conditions by the proportion of lost load. The specific calculation method is as follows: ； The economic indicators characterize the utilization efficiency of power grid equipment through the average load rate of the equipment, and the specific calculation method is as follows: ； The balanced group comprehensive evaluation calculates a comprehensive evaluation score by weighting and summing four categories of indicators: sufficiency, safety, cleanliness, and economy. The specific calculation method is as follows: ； in, The standardized scores are for the four categories of indicators; These are the weighting coefficients.

8. The hierarchical and autonomous balancing method for active distribution networks based on scenario-based multi-objectives as described in claim 7, characterized in that, The implementation steps of the differentiated active balancing strategy include: The global power balance adopts a hierarchical power flow optimization algorithm, which combines the resource endowment of the 10kV unit layer, the 110kV grid layer and the 220kV partition layer to optimize the cross-level power flow direction and clarify the annual construction scale of each voltage level. Regional scenario-based balancing addresses local bottleneck issues by filtering candidate measures through scenario matching rules and employing a cost-benefit priority algorithm to select the optimal solution and formulate specific optimization strategies. The 5G collaborative communication channel utilizes its millisecond-level low latency to enable real-time interaction of source and load data and balancing commands between the overall coordination center and local area nodes, ensuring the efficiency of information transmission. Establish a "global coordination - regional response" linkage mechanism, which coordinates cross-regional resource allocation at the global level and enables regions to respond quickly based on global instructions or their own scenario needs, through a closed-loop process of "intra-level optimization - inter-level power interaction - global optimal verification".

9. The scenario-based multi-objective active distribution network hierarchical autonomous balancing method according to claim 7, characterized in that, Adjustments are triggered when evaluation indicators fail to reach preset thresholds, i.e., when the loss load ratio is >0.5%, balancing measures are adjusted, and when the self-balancing coefficient is <0.8, group division is adjusted. The specific adjustment methods are as follows: when adjusting the group division, the source-load density ratio is recalculated and adjacent receiving and sending groups are merged to optimize the boundary; when adjusting the scenario characterization, the historical photovoltaic output data of the past year is supplemented to regenerate the typical daily curve; when adjusting the balancing measures, high-cost chemical energy storage is replaced with demand response and load coupling. The annually updated data includes load forecast data, newly installed capacity of new energy sources, power grid topology changes, and equipment aging parameters.

10. The scenario-based multi-objective active distribution network hierarchical autonomous balancing method according to claim 9, characterized in that, The closed-loop optimization cycle is to carry out a full-process iterative optimization once a year, which means repeating the process of group division, scenario characterization, active balancing, and state evaluation to dynamically adjust the balancing strategy and the scale of power grid construction.

Images