Integrated research and judgment method for active power distribution network planning and regulation based on joint optimization model

Abstract

This invention discloses an integrated analysis method for planning and regulation of active distribution networks based on a joint optimization model, relating to the fields of power systems and their automation. The method includes: collecting multi-source data from the active distribution network and generating an integrated data pool; constructing an operational situation awareness model to perceive multi-dimensional situation information in real time and outputting analysis results through an intelligent analysis model; constructing a joint optimization model based on the analysis results, comprising an objective function layer, a decision variable layer, a constraint condition layer, and a coupling constraint layer, wherein the coupling constraint layer represents the bidirectional coupling relationship between planning and regulation decisions; employing a hybrid solution strategy combining mixed integer programming and model predictive control for solving the problem, achieving coordinated convergence through iterative interaction, and outputting the optimal integrated decision scheme; and performing closed-loop feedback and dynamic rolling updates with preset periods and / or preset trigger conditions. This invention achieves integrated coordination and dynamic optimization of active distribution network planning and regulation.

Description

Technical Field

[0001] This invention relates to the field of power systems and their automation technology, specifically to an integrated analysis method for planning and regulation of active distribution networks based on a joint optimization model. Background Technology

[0002] An active distribution network is a modern power system architecture that, unlike traditional passive distribution networks, possesses the ability to actively regulate, control, and manage the flow of electrical energy. This system integrates various energy resources and advanced energy management technologies, such as solar photovoltaic, wind power, energy storage systems, and intelligent control and forecasting algorithms. Active distribution networks can adjust the generation, storage, and consumption of energy in real time according to electricity demand and supply conditions to optimize energy use efficiency, and to supply power to or receive power from the grid when necessary.

[0003] With the large-scale integration of distributed power sources (such as photovoltaic and wind power), energy storage systems, electric vehicle charging facilities, and flexible loads, traditional passive distribution networks are gradually evolving into active distribution networks. The flexibility and intelligence of active distribution networks enable them to more stably and reliably meet the ever-increasing energy demand, while promoting the integration and utilization of renewable energy and driving the power system towards a more sustainable and environmentally friendly direction.

[0004] Currently, the planning and operation regulation of active distribution networks typically adopt a phased and sequential processing model at the technical practice level. Distribution network planning mainly determines the grid structure, distributed power source location and capacity, and energy storage configuration schemes on a medium- to long-term time scale; while operation regulation, on a short-term or real-time time scale, performs power flow optimization, voltage regulation, and power dispatch based on the existing grid structure. There is a lack of effective two-way information feedback and collaborative optimization mechanisms between the two.

[0005] For example, Chinese invention patent CN104866924A discloses a joint optimization method for active distribution network planning and operation. This method includes listing decision variables based on alternative planning schemes, listing objective functions based on planning objectives, listing planning constraints and operational constraints, generating a joint optimization model for active distribution network planning and operation based on the objective function and constraints, and calculating the optimization model to obtain the optimal decision. Another example is Chinese invention patent CN112581039A, which discloses a joint optimization method for active distribution network planning and operation. This method first obtains distribution network planning data for the target area through planning decisions, establishes a distribution network model and solves it to obtain planning decision schemes, then obtains planning decision schemes through operational decisions and analyzes operational and profitability conditions to derive operational decision schemes. Yet another example is Chinese invention patent CN120222468B, which discloses a source-load coordinated planning method for active distribution networks considering optimal power flow. This method establishes a source-load coordinated planning objective function and constraints considering optimal power flow, and uses optimization algorithms to perform multi-objective optimization of the optimal location and allocation ratio of distributed power sources connected to the distribution network.

[0006] However, the aforementioned existing technologies still have the following shortcomings:

[0007] (i) Planning and regulation are separated, lacking an integrated and coordinated mechanism. Most existing methods treat the generation of planning schemes and real-time regulation decisions as two sequential stages. Once the planning scheme is determined, it lacks the ability to dynamically adjust based on real-time operating conditions, making it difficult to cope with the strong uncertainty of distributed power output and rapid load fluctuations. Although some methods have proposed "joint optimization" of planning and operation, it is essentially still a sequential mode of "planning first, operation following," lacking a two-way information interaction and iterative feedback mechanism between the planning layer and the operation layer.

[0008] (ii) Lack of an intelligent situation-driven assessment mechanism. Existing technologies have failed to construct a closed-loop integrated framework from "data perception → situation assessment → planning optimization → adjustment decision-making". Planners and dispatchers rely on their respective tools and models to make decisions separately, lacking a unified data foundation and intelligent situation assessment methods, resulting in low decision-making efficiency, poor coordination, and difficulty in achieving dynamic adaptation between planning schemes and operational strategies.

[0009] (iii) The high model complexity makes it difficult to simultaneously meet the multi-timescale requirements of long-term planning and real-time adjustment. Traditional mixed-integer programming methods are computationally complex and therefore cannot directly support online decision-making; while purely intelligent optimization methods have limited generalization ability when faced with scenarios such as topology changes and equipment deployment / removal. Existing technologies lack a multi-timescale hierarchical solution strategy that can effectively connect long-term planning decisions with short-term operational adjustments.

[0010] (iv) Lack of closed-loop dynamic optimization capability. Most existing methods are static planning. Once the planning scheme is determined, it remains unchanged for a long period of time. It cannot be continuously optimized and dynamically adjusted according to the evolution of the actual operating status of the distribution network, which leads to the cumulative increase of the deviation between the planning scheme and the actual operating needs over time.

[0011] To address the shortcomings of the existing technologies, this invention provides an integrated analysis method for active distribution network planning and regulation based on a joint optimization model. Summary of the Invention

[0012] The technical problem to be solved by the present invention is to overcome the existing defects and provide an integrated judgment method for active distribution network planning and regulation based on a joint optimization model, which can realize the coordinated optimization and dynamic updating of planning schemes and regulation strategies, and can effectively solve the problems in the background technology.

[0013] To achieve the above objectives, the present invention provides the following technical solution: an integrated analysis method for active distribution network planning and regulation based on a joint optimization model, comprising the following steps:

[0014] S1: Collect multi-source data from the active power distribution network, perform data cleaning, normalization, and spatiotemporal alignment on the multi-source data, and generate a standardized integrated data pool.

[0015] Specifically, the multi-source data in step S1 includes: power grid topology data, equipment parameter data, historical operating data, distributed power generation output forecast data, load forecast data, meteorological data, real-time switch status data, and energy storage system operating status data. Spatiotemporal alignment processing involves mapping data with different time scales and spatial granularities to a common timestamp and spatial node.

[0016] The data includes: power grid topology data (connection relationships and electrical parameters of components such as buses, lines, and transformers in the distribution network); equipment parameter data (rated capacity, impedance parameters, and operating limits of various power equipment); historical operating data (node ​​voltage, branch power flow, and equipment load rate records at past time points); distributed power generation output forecast data (active and reactive power output forecast curves for distributed power sources such as photovoltaics and wind power in future periods); load forecast data (power demand forecast curves for each load node in future periods); meteorological data (meteorological factors such as irradiance, wind speed, temperature, and humidity that affect the output and load levels of distributed power sources); real-time switch status data (current open / closed status of circuit breakers and disconnectors in the distribution network); and energy storage system operating status data (current state of charge and maximum charge / discharge power of energy storage devices).

[0017] Data cleaning is used to remove outliers and missing values ​​from multi-source data. Normalization is used to map data of different dimensions to a unified numerical range. Spatiotemporal alignment is used to map data from different acquisition systems with different temporal resolutions and spatial granularities to a common timestamp and spatial node, forming a standardized integrated data pool that provides a unified data foundation for subsequent steps.

[0018] S2: Based on an integrated data pool, an active power distribution network operation status awareness model is constructed to perceive multi-dimensional status information in real time, and an intelligent judgment model is used to comprehensively judge the multi-dimensional status information and output the judgment results.

[0019] Specifically, the multi-dimensional situational information in step S2 includes: node voltage distribution, branch power flow distribution, distributed generation output fluctuation, load change, grid security margin, and renewable energy consumption. The assessment results include: grid operational health score, risk warning level, urgency of regulation needs, and recommendations for adjusting planning schemes.

[0020] Among them, the node voltage distribution trend reflects the current voltage level of each node in the distribution network and its proximity to the upper and lower voltage limits; the branch power flow distribution trend reflects the current load rate of each feeder and transformer and its proximity to the thermal stability limit; the distributed generation output fluctuation trend reflects the time change rate, fluctuation amplitude and prediction error of the distributed generation output; the load change trend reflects the time change rate of the load, the peak occurrence period and uncertainty range; the grid safety margin trend reflects the distance between the current operating point of the distribution network and the safe operating boundary; and the renewable energy consumption trend reflects the system's ability to consume and abandon distributed renewable energy generation.

[0021] The intelligent assessment model is a multi-task learning model built on a deep neural network. It takes time-series multi-dimensional situational information as input and outputs grid operation health score, risk warning level, urgency of regulation needs, and planning scheme adjustment suggestions in parallel. It is trained using historical operational data from the active distribution network. Based on the current multi-dimensional situational information, the model can automatically output a grid operation health score (used to quantitatively assess the overall quality of the current distribution network operation), a risk warning level (used to identify the level of current or impending operational risks), a regulation need urgency level (used to measure the urgency of current operational regulation), and planning scheme adjustment suggestions (used to indicate whether and how existing planning schemes need to be adjusted).

[0022] S3: Based on the analysis results, construct a joint optimization model that integrates planning and adjustment. The joint optimization model includes an objective function layer, a decision variable layer, a constraint condition layer, and a coupling constraint layer.

[0023] (I) Objective Function Layer. The objective function layer in step S3 aims to minimize overall cost. The objective function includes four sub-objectives: planning period investment and construction costs, operation and regulation costs, reliability costs, and renewable energy consumption benefits. Planning period investment and construction costs include line construction costs, substation expansion costs, and distributed power generation and energy storage installation costs. Operation and regulation costs include network loss costs, equipment regulation costs, and demand response compensation costs. Reliability costs include power outage loss costs and voltage over-limit penalty costs. Renewable energy consumption benefits are included in the objective function in the form of wind and solar curtailment penalty costs.

[0024] Each sub-objective is comprehensively optimized using a multi-objective processing method. Preferably, the weighted summation method is used, where each sub-objective is assigned a corresponding weight coefficient and then summed to obtain the comprehensive objective function. Alternatively, the ε-constraint method or the Pareto optimality method can also be used for multi-objective processing.

[0025] (II) Decision Variable Layer. The decision variable layer in step S3 includes two types of variables: planning decision variables and adjustment decision variables. Planning decision variables include line construction status, distributed power source location and capacity determination, energy storage system configuration capacity and power, and reactive power compensation device configuration. Adjustment decision variables include active and reactive power output of distributed power sources, energy storage charging and discharging power, adjustable transformer tap position, flexible load adjustment amount, and switch status.

[0026] Among them, the line construction status is a 0-1 variable, indicating whether each candidate line has been constructed; the distributed power source location and capacity setting are integer or continuous variables, representing the installed capacity of distributed power sources at each candidate access point; the energy storage system configuration capacity and power are continuous variables, representing the rated capacity and rated charging and discharging power of energy storage at each candidate installation point; the reactive power compensation device configuration is a discrete or continuous variable, representing the configuration capacity of parallel capacitors or reactors at each node. In the adjustment decision variables, the active and reactive power outputs of distributed power sources are continuous variables, representing the actual active and reactive power outputs of each distributed power source during operation; the energy storage charging and discharging power is a continuous variable, representing the charging and discharging power of energy storage at each time period; the adjustable transformer tap position is a discrete variable, representing the tap position of the on-load tap-changing transformer; the flexible load adjustment amount is a continuous variable, representing the adjustment amount of interruptible or transferable loads; and the switch status is a 0-1 variable, representing the open / closed status of each tie switch or sectionalizing switch.

[0027] (III) Constraint Layer. The constraint layer in step S3 includes planning constraints and operational constraints. Planning constraints include investment budget constraints, land resource constraints, equipment selection constraints, and grid N-1 security criterion constraints. Operational constraints include node power balance constraints, branch power flow constraints, node voltage upper and lower limit constraints, equipment output upper and lower limit constraints, energy storage state of charge constraints, and adjustable load adjustment range constraints.

[0028] Among these constraints, the investment budget constraint requires that the total investment and construction cost during the planning period not exceed the given budget ceiling; the land resource constraint requires that the installation locations of distributed power sources and energy storage meet land availability restrictions; the equipment selection constraint requires that the selected equipment types come from a given set of candidate equipment; the grid N-1 safety criterion constraint requires that the planned scheme can still maintain safe and stable operation of the system even if any single component is out of service. The node power balance constraint requires that the injected power and outflow power of each node remain balanced in each time period; the branch power flow constraint requires that the active and reactive power transmitted by each branch does not exceed its thermal stability limit; the node voltage upper and lower limit constraint requires that the voltage amplitude of each node remain within the allowable range; the equipment output upper and lower limit constraint requires that the output of various types of equipment does not exceed its rated capacity; the energy storage state of charge constraint requires that the state of charge of energy storage remain within a safe range during operation and meet the charging and discharging power limits; and the adjustable load adjustment range constraint requires that the adjustment amount of flexible loads be within their adjustable range.

[0029] (iv) Coupling Constraint Layer. The coupling constraint layer in step S3 is used to characterize the two-way coupling relationship between planning decisions and adjustment decisions. The two-way coupling relationship is as follows: the planning decision variables determine the type and capacity limit of the available adjustment resources, and the feasible domain of the adjustment decision variables changes dynamically with the changes in the planning decision variables; at the same time, the operational boundary conditions of the adjustment decision variables in various typical scenarios are fed back to the planning layer in the form of constraints, constituting additional constraints for the feasibility verification of the planning scheme.

[0030] The general mathematical form of the coupling constraint layer is: ,in For planning decision vectors, To adjust the decision vector.

[0031] Specifically, coupling constraints include, but are not limited to, the following forms: energy storage configuration capacity constraints regulating layer charging and discharging power—the charging and discharging power of energy storage in each time period shall not exceed the rated energy storage power determined by the planning layer; line construction status constraints branch power flow—if the line is not constructed, its transmission power must be 0, and if the line is constructed, its transmission power shall not exceed the thermal stability limit; distributed power generation installation capacity constraints its active power output—the actual output of distributed power generation in each time period shall not exceed the rated installation capacity determined by the planning layer.

[0032] Typical scenarios include high photovoltaic output scenarios, peak load scenarios, low photovoltaic output scenarios, and N-1 fault scenarios. The operational boundary conditions under each scenario are fed back to the planning layer in the form of constraints to ensure that the planning scheme is feasible in all typical scenarios.

[0033] S4: Solve the joint optimization model. Solve the planning layer decision variables on the first time scale and solve the adjustment layer decision variables based on the planning layer decision results on the second time scale. Achieve coordinated convergence of the planning layer and the adjustment layer through iterative interaction, and output the optimal integrated decision scheme that satisfies all constraints and minimizes the objective function.

[0034] In step S4, a hybrid solution strategy combining mixed-integer programming and model predictive control is used to solve the joint optimization model. The first time scale is annual, quarterly, or monthly, and mixed-integer linear programming or mixed-integer nonlinear programming is used to solve the planning-level decision variables. The second time scale is daily or hourly, and model predictive control is used to solve the adjustment-level decision variables, with the planning-level decision results serving as fixed parameter inputs for model predictive control.

[0035] The iterative interaction is as follows: the planning layer passes the decision results to the adjustment layer as the optimization boundary, and the adjustment layer feeds back the optimal operating cost under each typical scenario to the planning layer as a feedback value. This process is repeated alternately until the relative rate of change of the objective function is less than a preset convergence threshold. Preferably, the preset convergence threshold ranges from 1% to 5%.

[0036] The specific solution process is as follows: First, at the annual, quarterly, or monthly time scale, starting with the current initial values ​​of the planning decision variables, the planning layer sub-problems are solved using mixed-integer linear programming or mixed-integer nonlinear programming to obtain candidate planning schemes. Then, the planning decision variables in the candidate planning schemes are passed to the regulation layer as fixed parameters. Next, at the daily or hourly time scale, using the planning decision results as boundary conditions, the regulation layer sub-problems under each typical scenario are solved using model predictive control methods to obtain the optimal operation regulation strategy and the corresponding optimal operating cost under each scenario. Then, the optimal operating cost under each typical scenario is fed back to the planning layer as a weighted sum. The planning layer incorporates the feedback value into the objective function, resolves the planning layer sub-problems, and updates the planning decision variables. The above process is repeated until the relative rate of change of the objective function between two adjacent iterations is less than the preset convergence threshold, at which point co-convergence is determined, and the optimal integrated decision scheme is output.

[0037] S5: Based on the optimal integrated decision scheme output in step S4, generate and output an integrated decision scheme report for the active distribution network. The integrated decision scheme report includes the optimal network planning scheme, the optimized configuration scheme of distributed power sources and energy storage, and the operation and regulation strategy at multiple time scales.

[0038] The optimal network planning scheme includes a list of recommended new or expanded lines, their routes, and specifications; the optimized configuration scheme for distributed power sources and energy storage includes recommended values ​​for the installed capacity of distributed power sources at each candidate access point, and recommended values ​​for the rated capacity and rated power of energy storage at each candidate installation point; and the multi-timescale operation and regulation strategies include an annual-level planning scheme execution plan, a daily-level day-ahead scheduling plan, and an hourly to minute-level real-time regulation strategy.

[0039] S6: During actual operation, real-time operation data is continuously collected, and steps S1 to S5 are re-executed at a preset period and / or when preset trigger conditions are met, so as to realize the dynamic rolling optimization and update of planning schemes and adjustment strategies.

[0040] Specifically, the preset period can be monthly, quarterly, or annually, and can be set according to the actual scale and rate of change of the distribution network. Preset trigger conditions include, but are not limited to: node voltage exceeding limits for a set duration, branch load rate exceeding a set threshold for a set time, deviation between actual and predicted output of distributed power sources exceeding a set percentage for a set time, significant changes in system topology, energy storage system state of charge exceeding a safe range for a set time, and load level changes exceeding a set percentage. When the preset period is reached or a preset trigger condition is detected, steps S1 to S5 are automatically re-executed, enabling dynamic rolling optimization and updates of the planning scheme and adjustment strategy, ensuring that the planning scheme always adapts to the actual operating state of the distribution network.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] 1. This invention breaks down the barriers between traditional sequential planning and regulation decisions by constructing a joint optimization model that includes planning and regulation decision variables and setting a coupling constraint layer. It achieves bidirectional collaborative optimization of planning schemes and regulation strategies, overcoming the shortcomings of existing technologies that separate planning and operation. Compared to the sequential "planning first, operation follow" mode in existing technologies, this invention achieves collaborative convergence of the planning and regulation layers through an iterative interaction mechanism, resulting in a globally superior integrated decision scheme.

[0043] 2. This invention introduces an intelligent situational awareness and analysis step before joint optimization. By constructing a multi-dimensional situational awareness model and an intelligent analysis model, planning and adjustment decisions are based on a precise grasp of the real-time operational situation, realizing a shift from "passive response" to "active analysis," significantly improving the scientific rigor and timeliness of decision-making. Compared to the existing technology where planners and schedulers rely on their respective tools for decision-making, this invention achieves a closed-loop integrated decision-making process encompassing data awareness, situational analysis, planning optimization, and adjustment decisions.

[0044] 3. This invention adopts a hybrid solution strategy of long-term planning and short-term adjustment layered iteration—solving the planning layer with mixed integer programming at the annual, quarterly or monthly scale, and solving the adjustment layer with model predictive control at the daily or hourly scale—which greatly reduces the complexity of model solution while ensuring global optimality, and overcomes the technical problem that traditional methods are difficult to support online decision-making due to excessive computational complexity.

[0045] 4. This invention uses a closed-loop feedback mechanism to trigger the re-solution of planning and regulation at a preset period and / or when preset triggering conditions are met, thereby realizing the rolling optimization and updating of planning schemes and regulation strategies. This enables active distribution networks to effectively cope with the uncertainty of distributed power output and rapid load fluctuations, significantly improving the adaptability and robustness of the distribution network and overcoming the shortcomings of existing static planning methods that cannot dynamically adapt to actual operating needs. Attached Figure Description

[0046] Figure 1 This is a flowchart of the workflow of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Please see Figure 1 The present invention provides the following embodiments:

[0049] Example 1

[0050] Taking a specific active power distribution network area as an example, this invention's method is applied to conduct integrated planning and regulation analysis of the area. The basic situation of this area is as follows:

[0051] Voltage level: 10kV

[0052] Number of feeders: 3 10kV feeders

[0053] Distributed PV grid connection points: 5 candidate grid connection points, with a total existing installed capacity of 2MW.

[0054] Energy storage system access candidate sites: 2 candidate installation sites

[0055] Adjustable load: Several, total adjustable capacity approximately 1MW

[0056] Planning period: the next 5 years (2026-2030)

[0057] Perform the following steps according to the method of the present invention.

[0058] Step S1: Data Acquisition and Fusion Processing

[0059] Collect the following multi-source data for this region:

[0060] (1) Power grid topology data: bus-line connection relationship of the three feeders, electrical connection method between each node, and configuration location of each switchgear.

[0061] (2) Equipment parameter data: impedance parameters such as resistance and reactance of each line, thermal stability limit of each line i.e. current carrying capacity, rated capacity and short-circuit impedance of each distribution transformer, rated capacity and power factor adjustment range of each photovoltaic inverter, and rated power and charging and discharging efficiency of each energy storage converter.

[0062] (3) Historical operation data: node voltage amplitude and phase angle, active and reactive power flow of branches, actual output of each distributed power source, actual power of each load node, and operation records of each switching device at a sampling point every 15 minutes over the past 3 years (2023-2025).

[0063] (4) Distributed power generation output forecast data: the annual theoretical maximum output curves of each photovoltaic access point in the next 5 years, and the hourly output forecast data for the next week.

[0064] (5) Load forecast data: the annual maximum load forecast and growth rate of each load node in the next 5 years, and the hourly load curve of typical days (weekdays, weekends, holidays).

[0065] (6) Meteorological data: historical hourly irradiance, temperature and humidity data for the region, as well as hourly weather forecast data for the next week.

[0066] (7) Real-time switch status data: The current open / closed status of each tie switch and section switch is collected in real time through the SCADA system.

[0067] (8) Energy storage system operation status data: current state of charge, maximum chargeable power, maximum dischargeable power, battery health status, etc. of existing energy storage equipment.

[0068] The above multi-source data will be processed sequentially as follows:

[0069] Data cleaning: Outliers in each data source were removed using a statistical threshold-based method, and missing values ​​were filled using linear interpolation. For voltage data, the normal range was set to 85% to 115% of the rated voltage; data points outside this range were marked as outliers and removed. For data segments with no more than 5 consecutive missing sampling points, linear interpolation of the preceding and following valid values ​​was used to fill in the missing data.

[0070] Normalization: The min-max normalization method is used to map data of different dimensions to a unified numerical range. For voltage data, the rated voltage is used as the reference value for per-unit normalization; for power data, the system reference capacity is used as the reference value for per-unit normalization; for temperature data, the historical highest and lowest temperatures of the region are used as upper and lower limits for normalization.

[0071] Spatiotemporal alignment processing: Data with different temporal resolutions and spatial granularities are uniformly mapped to a common timestamp and spatial node. Specifically, annual planning data is discretized into hourly cross-sections in the time dimension, and the data of each node is spatially aligned according to the node number to generate a standardized time-series-spatial integrated data pool.

[0072] Step S2: Intelligent perception and analysis of operational status

[0073] Based on the integrated data pool obtained in step S1, an active distribution network operation status awareness model is constructed. In this embodiment, the status awareness model adopts a state estimation algorithm based on weighted least squares, using the measurement data (node ​​voltage amplitude, branch power, etc.) in the integrated data pool as input to estimate the current real operating state of the system (voltage amplitude and phase angle of each node).

[0074] Based on the state estimation results, the following six types of situation information are calculated in real time:

[0075] (1) Node voltage distribution: Calculate the degree of proximity of the voltage amplitude of each node to the upper and lower voltage limits, and express it quantitatively using the voltage margin index.

[0076] (2) Branch power flow distribution: Calculate the ratio of the current power flow of each branch to its thermal stability limit, and express it in terms of load rate.

[0077] (3) Distributed power output fluctuation trend: Calculate the rate of change of each distributed power output in the most recent hour and its percentage relative to the installed capacity.

[0078] (4) Load change trend: Calculate the rate of change of power of each load node in the most recent hour and its percentage relative to the maximum load.

[0079] (5) Power grid safety margin status: comprehensively calculate the static safety margin of the system, including voltage stability margin and power angle stability margin.

[0080] (6) New energy consumption status: Calculate the proportion of new energy power generation to the total system load at the current moment, and whether there is wind curtailment or solar curtailment.

[0081] The above six types of situation information are input into the intelligent assessment model. In this embodiment, the intelligent assessment model adopts a multi-task learning model based on deep neural networks. Its input layer is a one-dimensional convolutional layer used to extract the time-series features of the situation information; the hidden layer is a three-layer fully connected layer, each containing 128 neurons, and the activation function is a linear rectified function; the output layer consists of four parallel output heads, corresponding to the power grid operation health score (outputting a continuous value from 0 to 100), risk warning level (outputting a four-class probability), urgency of adjustment demand (outputting a three-class probability), and planning scheme adjustment suggestions (text generation task).

[0082] The model was trained using historical operational data from the region over the past three years. The training data was labeled as follows: three power system dispatching experts with intermediate or higher professional titles independently labeled each historical sample, and the majority consistent label was used as the final label. The training objectives were to minimize the mean squared error of the health score and the cross-entropy loss of the risk warning level and the urgency of the adjustment demand.

[0083] In this embodiment, the current grid operation health score output by the intelligent assessment model is 82 points (out of 100), the risk warning level is "general risk", the urgency of adjustment demand is "adjustment recommended", and the planning scheme adjustment suggestion is: "The output of photovoltaic access point 3 is fluctuating significantly, reaching 22% of the rated output within 15 minutes. It is recommended to configure energy storage near access point 3 to smooth out the output fluctuation; at the same time, the voltage of access point 5 is close to the upper limit, reaching 1.065 times the rated voltage. It is recommended to configure a reactive power compensation device at access point 5 or adjust the transformer tap."

[0084] Step S3: Construct a joint optimization model

[0085] Based on the analysis results of step S2, a joint optimization model integrating planning and regulation is constructed. In this embodiment, the planning period is set to 5 years, with a planning time step of 1 year; the time resolution of operation regulation is 1 hour, and four typical days (high photovoltaic output day, peak load day, low photovoltaic output day, and normal operation day) represent the typical operation scenarios of each season, with 24 time sections taken for each typical day.

[0086] (I) Objective Function Layer

[0087] With the goal of minimizing overall cost, the objective function contains four sub-objectives:

[0088] (1) Investment and construction costs during the planning period: including the cost of new construction or expansion of candidate lines, which is calculated based on the line length and cross-sectional specifications; the cost of expansion installation of photovoltaic access points, which is the unit capacity installation cost multiplied by the expansion capacity; and the cost of energy storage system configuration at energy storage candidate points, which includes capacity cost and power cost.

[0089] (2) Operation and adjustment costs: including network loss costs during the planning period, which are calculated based on the power flow calculation results under each typical scenario; equipment adjustment costs, which are maintenance costs generated by transformer tapping operations, energy storage charging and discharging switching, etc.; and demand response compensation costs, which are compensation fees paid to users when adjustable loads are called.

[0090] (3) Reliability cost: This includes the cost of power outage losses caused by power outage accidents, which is calculated based on the load importance of each node and the expected power outage time; and the cost of voltage over-limit penalty, which is the corresponding penalty when the node voltage exceeds the allowable range.

[0091] (4) Benefits of new energy consumption: The cost of curtailment of wind and solar power is included in the objective function. When the system cannot fully consume photovoltaic power generation, the curtailed solar power is multiplied by the penalty price coefficient and included in the objective function to incentivize planning schemes and adjustment strategies to maximize the consumption of new energy.

[0092] Each sub-objective is comprehensively optimized using a weighted summation method. Based on the region's operational objectives and management policies, the weighting coefficients in this embodiment are set as follows: planning period investment and construction cost weight 0.35, operation and adjustment cost weight 0.25, reliability cost weight 0.20, and wind and solar curtailment penalty cost weight 0.20.

[0093] (ii) Decision variable layer

[0094] The planning decision variables include: the construction status of candidate lines, which is a 0-1 variable; the expansion capacity of photovoltaic access points, which is a continuous variable ranging from 0 to 2 MW per point; the rated energy storage capacity of energy storage candidate points, which is a continuous variable ranging from 0 to 4 MWh per point; the rated energy storage power, which is a continuous variable ranging from 0 to 2 MW per point; and the reactive power compensation device configuration capacity of each node, which is a discrete variable.

[0095] The adjustment decision variables include: the active and reactive power output of each distributed power source in each time period; the charging and discharging power of each energy storage in each time period; the position of the adjustable transformer tap in each time period; the adjustment amount of each flexible load in each time period; and the opening and closing status of each tie switch in each time period.

[0096] (III) Constraint Layer

[0097] Planning constraints include: total investment budget constraint, i.e., the total investment construction cost during the planning period shall not exceed the given budget limit; land resource constraint, i.e., the installation location of distributed power sources and energy storage must meet land availability restrictions, with 0-1 variables representing the land availability status of each candidate point; equipment selection constraint, i.e., the selected equipment type comes from a given set of candidate equipment; and grid N-1 safety criterion constraint, i.e., the system can still maintain safe and stable operation even if any single component is out of service.

[0098] Operational constraints include: power balance constraints for each node at each time period; power flow constraints for each branch at each time period, i.e., the transmitted power does not exceed its thermal stability limit; voltage constraints for each node, i.e., the voltage amplitude is maintained within the range of 93% to 107% of the rated voltage; output constraints for each device, i.e., it does not exceed its rated capacity; state of charge constraints for energy storage, i.e., it is maintained within the safe range of 20% to 90% during operation; and adjustable load adjustment range constraints, i.e., the adjustment amount does not exceed the upper and lower limits of its adjustable capacity.

[0099] (iv) Coupling Constraint Layer

[0100] The coupling constraint layer represents the two-way coupling relationship between planning decisions and adjustment decisions:

[0101] Energy storage configuration capacity constraints regulate the charging and discharging power of the adjustment layer—the charging and discharging power of energy storage in each time period shall not exceed the rated energy storage power determined by the planning layer. Line construction status constraints constrain branch power flow—if a line is not in operation, its transmission power must be 0; if a line is in operation, its transmission power shall not exceed the thermal stability limit. Distributed power generation installation capacity constraints constrain its active power output—the actual output of distributed power generation in each time period shall not exceed the rated installed capacity determined by the planning layer.

[0102] Meanwhile, the operational boundary conditions under each typical scenario (high photovoltaic output day, peak load day, low photovoltaic output day, and N-1 fault scenario) are fed back to the planning layer in the form of constraints to ensure that the planning scheme is feasible under all typical scenarios.

[0103] Step S4: Solving the joint optimization model

[0104] A hybrid solution strategy combining mixed-integer programming and model predictive control is adopted. In this embodiment, the mixed-integer linear programming problem is solved by calling the commercial solver Gurobi 10.0 using Python, and the model predictive control problem is solved using the MPC toolbox in MATLAB / Simulink.

[0105] At the annual scale, mixed-integer linear programming is used to solve for the planning-level decision variables, including line construction status, photovoltaic expansion capacity, energy storage configuration capacity and power, and reactive power compensation configuration. At the daily scale, based on the planning-level decision results, model predictive control is used to solve for the regulation-level decision variables, including the output of each distributed power source, energy storage charging and discharging power, transformer tap positions, flexible load regulation, and switch status. The planning-level decision results (such as rated energy storage power, line construction status, and photovoltaic installation capacity) serve as fixed parameter inputs for model predictive control. The prediction time domain for the regulation layer is 24 hours, the control time domain is 1 hour, and the rolling step size is 1 hour.

[0106] The iterative interaction process is as follows:

[0107] The first iteration: Starting with the initial planning scheme (no new lines, no expansion of photovoltaic capacity, and no energy storage configuration), the planning layer solves a mixed-integer linear programming problem to obtain candidate planning scheme A: It recommends building one new line (line 1, length 3.2 km, cross-section 185 mm²), expanding the photovoltaic capacity by 0.5 MW at photovoltaic access point 3, and configuring 1 MW / 2 MWh energy storage at energy storage candidate point 1. The planning decision variables of scheme A are passed to the regulation layer. Based on the decision results of scheme A, the regulation layer solves the MPC problem under four typical scenarios to obtain the optimal operation regulation strategy and optimal operating cost for each scenario. The weighted average operating cost is 3.2 million yuan per year. The weighted average operating cost is then fed back to the planning layer as feedback.

[0108] In the second iteration, the planning layer incorporates the feedback operating costs into the objective function and resolves the planning layer sub-problems, resulting in candidate planning scheme B: It recommends constructing two new lines (Line 1 and Line 2), expanding the photovoltaic grid connection at points 3 and 5 by 1 MW each, and configuring 1.5 MW / 3 MWh energy storage at energy storage candidate points 1 and 2 each. Based on scheme B, the regulation layer resolves the MPC problems under various typical scenarios, reducing the weighted average operating cost to 2.75 million yuan per year.

[0109] The third iteration: The planning layer solves again, obtaining candidate planning scheme C: It recommends constructing 2 new lines, expanding the photovoltaic grid connection point 3 by 1.2 MW, expanding the grid connection point 5 by 0.8 MW, configuring 2 MW / 4 MWh energy storage at candidate energy storage point 1, and configuring 1 MW / 2 MWh energy storage at candidate point 2. The regulation layer solves based on scheme C, with a weighted average operating cost of 2.68 million yuan per year.

[0110] Fourth iteration: The solution results of the planning layer are basically the same as those of the third iteration. The relative rate of change of the objective function is less than the preset convergence threshold of 3%, and the co-convergence is determined. The optimal integrated decision scheme (scheme C) is output.

[0111] Step S5: Generate an integrated decision-making solution report

[0112] Based on the solution results of step S4, an integrated decision-making scheme report is generated and output. This report is presented to distribution network planners and dispatching operators in the form of a visual dashboard. The report includes the following:

[0113] (1) Optimal grid planning scheme: It is recommended to build two new 10kV lines—Line 1 runs from A to B, with a length of 3.2 km and a conductor cross-section of 185 mm²; Line 2 runs from C to D, with a length of 2.8 km and a conductor cross-section of 185 mm². The construction sequence of the lines is as follows: Line 1 will be constructed in the first year, and Line 2 will be constructed in the third year. The total investment of the scheme is approximately RMB 32 million, which is within the budget of RMB 50 million.

[0114] (2) Optimized configuration scheme of distributed power source and energy storage: Expand the photovoltaic access point 3 by 1.2 MW to bring the total installed capacity to 1.6 MW; expand the photovoltaic access point 5 by 0.8 MW to bring the total installed capacity to 1.2 MW; configure a 2 MW / 4 MWh lithium iron phosphate energy storage system at energy storage candidate point 1, and configure a 1 MW / 2 MWh lithium iron phosphate energy storage system at energy storage candidate point 2. The total power of the energy storage system is 3 MW, the total capacity is 6 MWh, and the total investment is approximately RMB 12 million.

[0115] (3) Multi-timescale operation and regulation strategy: at the annual level, a five-year phased construction plan is formulated; at the quarterly level, the energy storage charging and discharging strategy is adjusted according to seasonal characteristics, with the focus on absorbing photovoltaic power in summer and peak shaving and valley filling in winter; at the daily level, a day-ahead scheduling plan for each typical day is formulated, including hourly distributed power output plan, energy storage charging and discharging plan, and transformer tap position plan; at the hour to minute level, a real-time rolling regulation strategy based on model predictive control is formulated.

[0116] The comprehensive evaluation of the scheme shows that, compared with the traditional sequential approach (planning first, then determining the operation strategy), the overall cost of this scheme is reduced by about 12.3%, the renewable energy consumption rate is increased from 78.5% to 87.2%, an increase of about 8.7 percentage points, and the node voltage qualification rate is increased from 94.3% to 99.1%.

[0117] Step S6: Closed-loop feedback and dynamic update

[0118] During actual operation, real-time operational data is continuously collected. Steps S1 to S5 are re-executed on a quarterly basis. Simultaneously, re-execution is triggered immediately when the following preset trigger conditions are detected: voltage exceeding limits for any node for more than 15 minutes; load rate exceeding 90% for more than 30 minutes; actual output of distributed power generation deviating from predicted output by more than 20% for more than 1 hour; significant changes in system topology, such as the addition of distributed power generation to the grid, maintenance shutdown of important lines, or the addition of new loads; state of charge of the energy storage system continuously below 20% or above 90% for more than 2 hours; load level changing by more than 15% compared to the same period last year.

[0119] In this embodiment, in the third year of operation, a large industrial and commercial load user (2 MW of new load) was added to the area, triggering a significant change in the load level in the preset trigger conditions. The system automatically detected this change and immediately triggered the re-execution of steps S1 to S5. After re-execution, the intelligent judgment model identified that the new load caused insufficient power supply capacity in the area. Based on the remaining planning period and the demand of the new load, the joint optimization model generated an adjustment plan adapted to the new load—adding a feeder to supply power to the user based on the original plan, and adjusting the configuration capacity of energy storage candidate point 2 from 1 MW / 2 MWh to 1.5 MW / 3 MWh to enhance peak-shaving capacity. The total investment increased by approximately 8 million yuan after the adjustment, but it is still within the remaining budget.

[0120] Through the aforementioned closed-loop feedback mechanism, the planning scheme and regulation strategy always maintain optimal adaptation to the actual operating state of the distribution network, effectively addressing the uncertainty of distributed power output and rapid load fluctuations.

[0121] Example 2

[0122] The main difference between this embodiment and Embodiment 1 lies in the selection of multi-objective processing methods and solution strategies, making it suitable for scenarios with higher requirements for solution accuracy and sufficient computing resources.

[0123] In terms of multi-objective processing methods, this embodiment uses the ε-constraint method instead of the weighted summation method in Embodiment 1. Specifically, the investment and construction cost during the planning period is minimized as the primary objective, while operation and regulation costs, reliability costs, and wind and solar curtailment penalty costs are used as constraints, each with its upper limit value. By gradually adjusting the values ​​of each constraint upper limit, a series of Pareto optimal solution sets are generated, allowing decision-makers to select the final solution from the solution set based on actual operational preferences. The final solution selected in this embodiment differs slightly from Solution C in Embodiment 1—to further reduce investment and construction costs, this solution reduces the configuration of energy storage candidate point 2 from 1 MW / 2 MWh to 0.5 MW / 1 MWh, and correspondingly increases the capacity of photovoltaic access point 5 to 1.2 MW, utilizing photovoltaic power generation to replace part of the energy storage function. The investment and construction cost of this solution is about 6% lower than that of Embodiment 1, but the operation and regulation cost is about 4% higher, providing decision-makers with different cost trade-off options.

[0124] In terms of solution strategy, this embodiment uses mixed-integer nonlinear programming instead of mixed-integer linear programming at the planning layer to more accurately handle nonlinear power flow constraints and nonlinear cost functions, such as the nonlinear relationship between energy storage lifetime degradation cost and charge / discharge depth. The mixed-integer nonlinear programming problem is solved using an external approximation algorithm, iterating alternately between inner-layer nonlinear programming to solve continuous variables and outer-layer mixed-integer programming to solve discrete variables. At the regulation layer, model predictive control is still used, with the prediction time domain extended to 48 hours to better capture the diurnal periodicity of load and distributed power output.

[0125] The remaining steps are the same as in Example 1, and will not be repeated here.

[0126] Example 3

[0127] The difference between this embodiment and Embodiment 1 lies in the setting of preset triggering conditions and the specific implementation of the intelligent judgment model. It is suitable for scenarios with higher requirements for intelligence level and better data collection conditions.

[0128] Regarding the preset triggering conditions, this embodiment adds the following triggering conditions based on Embodiment 1: the health status of the energy storage system decreases by more than 5 percentage points compared to the previous quarter; the daily power generation of distributed power sources changes by more than 30% compared to the previous day, applicable to the rapid switching between clear and cloudy weather in high photovoltaic penetration scenarios; a permanent fault occurs in a feeder of the distribution network, causing irreversible changes in the topology; significant changes occur in the electricity market price signal, such as a deviation of more than 20% between the real-time electricity price and the current forecast price; meteorological warning information indicates that extreme weather events, such as typhoons or blizzards, will occur within the next 24 hours. The diverse triggering conditions ensure that the system can promptly trigger the re-optimization of the planning scheme under various abnormal and changing scenarios, improving the system's adaptability and risk resistance.

[0129] Regarding the intelligent judgment model, this embodiment uses a deep learning model based on Long Short-Term Memory (LSTM) networks instead of the deep neural network model in Embodiment 1. LSTM networks are better able to capture long-term dependencies in time-series data and are suitable for modeling the temporal evolution patterns of operational trends.

[0130] The structure of the Long Short-Term Memory (LSTM) network model is as follows: The input layer receives time-series data from an integrated data pool over the past 168 hours, including 86 time-series features such as voltage amplitude of each node, active and reactive power flow of each branch, output of each distributed power source, power of each load, state of charge of each energy storage unit, and state of each switch; the hidden layer is a two-layer LSTM network, each containing 256 memory units; the discard layer is placed between the two LSTM network layers with a discard rate of 0.2 to prevent overfitting; the output layer is a fully connected layer that outputs the current grid health score, risk warning level, urgency of regulation needs, and planning scheme adjustment suggestions.

[0131] The model was trained using historical operational data from the region over the past five years. It employed an adaptive moment estimation optimizer with a learning rate of 0.001, a batch size of 64, a maximum training epoch of 200 epochs, and a patience value of 20 epochs for the early stop strategy (meaning training stopped if the validation set loss did not decrease for 20 consecutive epochs). The training data was labeled by three senior power system experts. Testing showed that this Long Short-Term Memory (LSTM) network model achieved a 92.3% accuracy rate in assessing operational risks, approximately 2.6 percentage points higher than the deep neural network model in Example 1, and performed particularly well in handling long-term sequential dependencies.

[0132] The remaining steps are the same as in Example 1, and will not be repeated here.

[0133] Example 4

[0134] The difference between this embodiment and Embodiment 1 is that the adjustment layer in the solution strategy uses mixed integer model predictive control instead of conventional model predictive control, which is suitable for scenarios where the adjustment decision variables include discrete switching variables and where higher control accuracy is required.

[0135] Conventional model predictive control typically employs relaxation methods when dealing with discrete variables, which involves relaxing 0-1 variables to continuous variables and then rounding them to the nearest integer. This method may introduce some control error. In this embodiment, the optimization problem of the regulation layer is essentially a mixed-integer quadratic programming problem because the objective function includes a quadratic term for the power regulation, and the decision variables include discrete transformer tap positions and switching states.

[0136] The mixed-integer model predictive control directly solves the mixed-integer quadratic programming problem without relaxation in each rolling time domain, using a branch-and-bound method. The maximum solution time is set to 60 seconds, and the maximum number of branch nodes is 10,000. While ensuring solution accuracy, computational complexity is controlled by appropriately setting the prediction and control time domains. Experimental results show that the single-run solution time of the mixed-integer model predictive control is approximately 8 to 15 seconds, which meets the time requirements of online rolling optimization.

[0137] After adopting mixed-integer model predictive control, the switching states and transformer tap positions of the regulation layer decisions no longer require post-processing rounding, improving the control accuracy by approximately 3.5% compared to conventional model predictive control. This embodiment is particularly suitable for application scenarios in distribution networks where switching equipment adjustments are frequent and transformer tap actions have a significant impact on voltage quality.

[0138] The remaining steps are the same as in Example 1, and will not be repeated here.

[0139] Example 5

[0140] The difference between this embodiment and Embodiment 1 lies in the specific implementation of the coupling constraint layer, and it is applicable to active distribution network scenarios with a high proportion of distributed power sources.

[0141] In this embodiment, in addition to the energy storage configuration constraints, line construction constraints, and distributed power generation capacity constraints included in Embodiment 1, the coupling constraint layer also includes the following coupling constraints:

[0142] (1) Reactive power output of the reactive power compensation configuration constraint adjustment layer: The capacity of the static var generator configured at each node constrains the reactive power injection of the node in each time period, that is, the actual reactive power injection in each time period shall not exceed the rated capacity of the static var generator and shall not be less than the negative rated capacity.

[0143] (2) Transformer capacity constraint for load transfer: The rated capacity of the adjustable transformer constrains the power transmission through the transformer in each time period, that is, the power transmitted in each time period does not exceed the rated capacity of the transformer.

[0144] (3) Multi-scenario coupling constraints: The feasibility conditions of the adjustment layer under each typical scenario are packaged into additional constraints of the planning layer to ensure that the planning scheme has a feasible adjustment strategy in any scenario. Specifically, for high photovoltaic output scenarios, the system is required to have sufficient downward adjustment capability to absorb the peak photovoltaic output; for N-1 fault scenarios, the system is required to be able to restore safe operation through adjustment means after losing any component.

[0145] Through the alternating transmission and verification of the aforementioned coupling constraints between the planning and regulation layers, the feasibility and optimality of the planning scheme are effectively guaranteed in all typical scenarios. Taking the active distribution network with a high proportion of distributed power sources in this embodiment as an example, the distributed photovoltaic penetration rate in this area has reached 60% of the peak load. By adding the aforementioned coupling constraints, the voltage qualification rate of the planning scheme under the N-1 fault scenario is increased from 85.6% to 97.2%, effectively ensuring power supply reliability.

[0146] The remaining steps are the same as in Example 1, and will not be repeated here.

[0147] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for integrated analysis of active distribution network planning and regulation based on a joint optimization model, characterized in that, Includes the following steps: S1: Collect multi-source data from the active power distribution network, perform data cleaning, normalization and spatiotemporal alignment on the multi-source data, and generate a standardized integrated data pool. S2: Based on an integrated data pool, construct an active distribution network operation status awareness model to perceive multi-dimensional status information in real time, and use an intelligent judgment model to comprehensively judge the multi-dimensional status information and output the judgment results. S3: Based on the analysis results, construct a joint optimization model that integrates planning and adjustment. The joint optimization model includes an objective function layer, a decision variable layer, a constraint condition layer, and a coupling constraint layer. S4: Solve the joint optimization model. Solve the decision variables of the planning layer at the first time scale, and solve the decision variables of the adjustment layer based on the decision results of the planning layer at the second time scale. Achieve coordinated convergence of the planning layer and the adjustment layer through iterative interaction, and output the optimal integrated decision scheme that satisfies all constraints and minimizes the objective function. S5: Based on the optimal integrated decision scheme output in step S4, generate and output an integrated decision scheme report for the active distribution network. The integrated decision scheme report includes the optimal network planning scheme, the optimized configuration scheme of distributed power sources and energy storage, and the operation and regulation strategy at multiple time scales. S6: During actual operation, real-time operation data is continuously collected, and steps S1 to S5 are re-executed at a preset period and / or when preset trigger conditions are met, so as to realize the dynamic rolling optimization and update of planning schemes and adjustment strategies.

2. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model as described in claim 1, characterized in that: The multi-source data in step S1 includes: power grid topology data, equipment parameter data, historical operation data, distributed power generation output forecast data, load forecast data, meteorological data, real-time switch status data, and energy storage system operation status data; the spatiotemporal alignment processing is to uniformly map data with different time scales and spatial granularities to a common timestamp and spatial node.

3. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model as described in claim 1, characterized in that: The multi-dimensional situational information in step S2 includes: node voltage distribution, branch power flow distribution, distributed power generation output fluctuation, load change, grid safety margin, and renewable energy consumption. The assessment results include: grid operation health score, risk warning level, urgency of regulation demand, and planning scheme adjustment suggestions.

4. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 1, characterized in that: In step S3, the objective function layer aims to minimize the overall cost. The objective function includes four sub-objectives: planning period investment and construction cost, operation and regulation cost, reliability cost, and renewable energy consumption benefit. The planning period investment and construction cost includes line construction cost, substation expansion cost, and distributed power source and energy storage installation cost. The operation and regulation cost includes network loss cost, equipment regulation cost, and demand response compensation cost. The reliability cost includes power outage loss cost and voltage over-limit penalty cost. The renewable energy consumption benefit is included in the objective function in the form of wind and solar curtailment penalty cost.

5. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 1, characterized in that: In step S3, the decision variable layer includes two types of variables: planning decision variables and adjustment decision variables. Planning decision variables include line construction status, distributed power source location and capacity setting, energy storage system configuration capacity and power, and reactive power compensation device configuration. Adjustment decision variables include active and reactive power output of distributed power sources, energy storage charging and discharging power, adjustable transformer tap position, flexible load adjustment amount, and switch status.

6. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 1, characterized in that: In step S3, the constraint layer includes planning constraints and operational constraints. Planning constraints include investment budget constraints, land resource constraints, equipment selection constraints, and grid N-1 safety criterion constraints. Operational constraints include node power balance constraints, branch power flow constraints, node voltage upper and lower limit constraints, equipment output upper and lower limit constraints, energy storage state of charge constraints, and adjustable load adjustment range constraints.

7. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 1, characterized in that: In step S3, the coupling constraint layer is used to characterize the bidirectional coupling relationship between planning decisions and adjustment decisions. The bidirectional coupling relationship is as follows: planning decision variables determine the type and capacity limit of available adjustment resources; the feasible domain of adjustment decision variables changes dynamically with changes in planning decision variables; simultaneously, the operational boundary conditions of adjustment decision variables under various typical scenarios are fed back to the planning layer in the form of constraints, constituting additional constraints for verifying the feasibility of the planning scheme. The general mathematical form of the coupling constraint layer is: ,in For planning decision vectors, To adjust the decision vector.

8. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 1, characterized in that: In step S4, a hybrid solution strategy combining mixed integer programming and model predictive control is used to solve the joint optimization model. The first time scale is annual, quarterly, or monthly, and mixed integer linear programming or mixed integer nonlinear programming is used to solve the decision variables of the planning layer. The second time scale is daily or hourly, and model predictive control is used to solve the decision variables of the adjustment layer, where the decision results of the planning layer are used as fixed parameter inputs for model predictive control. The iterative interaction is as follows: the planning layer passes the decision results to the adjustment layer as the optimization boundary of the adjustment layer, and the adjustment layer feeds back the optimal operating cost under each typical scenario to the planning layer in the form of feedback values. The process is repeated alternately until the relative rate of change of the objective function is less than the preset convergence threshold.

9. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 4, characterized in that: Each sub-objective is comprehensively optimized using a multi-objective processing method; the multi-objective processing method can be any one of the weighted summation method, ε-constraint method, or Pareto optimality method.

10. The integrated analysis method for active distribution network planning and regulation based on a joint optimization model according to claim 8, characterized in that: The preset convergence threshold ranges from 1% to 5%.

Images