A method for generating a distributed photovoltaic substation energy storage optimization strategy

Abstract

The application discloses a kind of distributed photovoltaic substation energy storage optimization strategy generation method.The application realizes the dynamic weight distribution of multi-objective optimization model optimization goal by accurate determination of target substation operating scene category, makes optimization goal and the core operation risk of substation under different scene accurate adaptation, ensure that the direction of energy storage optimization is consistent with the actual operation demand of substation.At the same time in the iterative solution process of particle swarm optimization algorithm, abandon the traditional punishment elimination type constraint processing method, based on the sensitivity information of power grid calculated from power grid topological structure data, the candidate solution that violates the safe operation constraint of power grid is corrected directionally, which not only effectively improves the search efficiency of the algorithm under complex constraint conditions, avoids the algorithm into local optimum, but also can guarantee the feasibility of the solution obtained by iteration, and can maintain population diversity by retaining corrected particles, help the algorithm to solve better results.

Description

Technical Field

[0001] This invention relates to the technical field of energy storage optimization, and specifically to a method for generating an energy storage optimization strategy for distributed photovoltaic power stations. Background Technology

[0002] With the advancement of the "dual carbon" goals, distributed photovoltaic power, with its core advantages of "local development and local consumption," has achieved large-scale access in urban and rural power distribution network areas (especially county, township, and urban community power distribution network areas), and has become the mainstream model for the development of new energy in power distribution networks.

[0003] In related technologies, in the scenario of energy storage configuration and operation optimization in distributed photovoltaic power stations, existing solutions based on particle swarm optimization algorithms suffer from low search efficiency under complex constraints and the inability of optimization results to dynamically adapt to the operation scenario of the power station due to the use of a "penalty-elimination" constraint processing mechanism and fixed weight multi-objective configuration. Summary of the Invention

[0004] The technical problem this invention aims to solve is that, in the context of distributed photovoltaic (PV) power grid (PV) power grid (PV) power storage configuration and operation optimization, existing particle swarm optimization (PSO) algorithms suffer from low search efficiency under complex constraints and the inability to dynamically adapt optimization results to the PV power grid's operational scenario due to their "penalty-elimination" constraint processing mechanism and fixed-weight multi-objective configuration. The purpose is to provide a method for generating energy storage optimization strategies for distributed PV power grids, thereby solving the problems of low search efficiency and the inability to dynamically adapt optimization results to the PV power grid's operational scenario under complex constraints.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a method for generating a distributed photovoltaic (PV) area energy storage optimization strategy, comprising:

[0007] Acquire real-time operating data, power grid topology data, and photovoltaic output and load power forecast data for the target distribution area in the future preset time period;

[0008] Based on real-time operation data and forecast data, the operation scenario category of the target transformer area in the current and future preset time periods is determined, and the weights of each optimization objective in the preset multi-objective optimization model are dynamically assigned based on the operation scenario category; wherein, the multi-objective optimization model aims to minimize the number of voltage over-limits and minimize the system bus loss, and its constraints include energy storage body operation constraints and grid safety operation constraints.

[0009] The multi-objective optimization model is iteratively solved using a particle swarm optimization algorithm to obtain the optimized configuration parameters and operating strategy of the energy storage system. In each iteration, the following steps are performed: if a candidate solution corresponding to a particle in the population is determined to violate the grid safety operation constraints, the corresponding grid sensitivity information is calculated based on the grid topology data. Based on the power adjustment direction indicated by this grid sensitivity information, the decision variables of the particle are directionally corrected to generate a corrected particle that satisfies the grid safety operation constraints. The corrected particle is then reintroduced into the population to participate in subsequent iterative optimizations.

[0010] Furthermore, the step of acquiring real-time operating data of the target distribution area, grid topology data, and photovoltaic output and load power prediction data for a future preset period includes:

[0011] Through the data acquisition interface, the voltage, current, real-time photovoltaic output, and real-time load power of each monitoring point in the target area are continuously acquired as real-time operation data.

[0012] Obtain the power grid topology description data of the target transformer area. The power grid topology description data includes at least meter box location information, branch electrical parameters and transformer rated parameters. Based on the meter box location information and branch electrical parameters, determine the electrical connection relationship between nodes to form power grid topology structure data for calculation.

[0013] Historical operational data and future weather forecast data are acquired and processed through a pre-trained time series prediction model to obtain the initial photovoltaic output and load power curves for a preset future period. The initial photovoltaic output and load power curves are then smoothed and filtered to generate photovoltaic output and load power prediction data.

[0014] Furthermore, the step of determining the operating scenario category of the target transformer area in the current and future preset time periods based on real-time operating data and predicted data, and dynamically assigning weights to each optimization objective in the preset multi-objective optimization model based on the operating scenario category, includes:

[0015] Based on the real-time operating data, forecast data, and power grid topology data, at least three types of scenario characteristic indicators are calculated; among them, the scenario characteristic indicators include: a first indicator for characterizing power surplus and reverse power supply risk, a second indicator for characterizing the sensitivity of node voltage to power changes, and a third indicator for characterizing the severity of photovoltaic output or load power fluctuations.

[0016] Based on the values ​​of at least three types of scenario characteristic indicators and preset scenario determination rules, the current operating scenario category is determined; wherein, the operating scenario category includes at least the scenario of a combination of photovoltaic peak and load trough, a scenario of a combination of photovoltaic trough and load peak, a scenario of a combination of photovoltaic fluctuation and load stability, and a scenario of a combination of photovoltaic stability and random load.

[0017] Based on the preset weight configuration corresponding to the determined operating scenario category, weights are dynamically assigned to the two optimization objectives in the multi-objective optimization model: minimizing the number of voltage overruns and minimizing system bus losses. The weight configuration under different operating scenario categories is adapted to the core operating risks that need to be prioritized and controlled under that category.

[0018] Furthermore, the step of calculating at least three types of scenario characteristic indicators based on the real-time operating data, the predicted data, and the power grid topology data includes:

[0019] The remaining photovoltaic power is calculated based on the current and predicted photovoltaic output and load power. The first index is calculated by combining the rated active power capacity of the transformer and the distance correction factor determined based on the electrical distance between the photovoltaic injection point and the transformer outlet in the electrical topology.

[0020] Based on the power grid topology data, the sensitivity coefficients of the voltage of selected nodes in the distribution area to active power injection and reactive power injection are calculated, and the two sensitivity coefficients are combined according to a preset ratio to obtain the second index.

[0021] Based on the standard deviation and average power of photovoltaic output and load power within a preset time window, the coefficient of variation of the two are calculated respectively. After normalization by the historical maximum coefficient of variation, the maximum value is taken to obtain the third index.

[0022] Furthermore, the step of determining the current operating scenario category based on the values ​​of at least three types of scenario feature indicators and preset scenario determination rules includes:

[0023] The value of the first indicator is compared with the first preset threshold, and the value of the third indicator is compared with the second preset threshold.

[0024] Based on the comparison results between the first indicator and the first preset threshold, it is determined that the current power balance state of the transformer area is either a power surplus state or a power deficit state.

[0025] Based on the comparison results between the third indicator and the second preset threshold, it is determined whether the current power fluctuation state of the transformer area belongs to a high fluctuation state or a low fluctuation state.

[0026] Based on the combination of power balance state and power fluctuation state, a preset operating scenario category is mapped. If the state is determined to be a power surplus state with low fluctuation, the operating scenario category is determined to be a combination of photovoltaic peak and load trough. If the state is determined to be a power deficit state with low fluctuation, the operating scenario category is determined to be a combination of photovoltaic trough and load peak. If the power fluctuation state is determined to be a high fluctuation state, the dominant source of fluctuation is distinguished based on the numerical relationship between the photovoltaic output fluctuation component and the load power fluctuation component, which constitute the third indicator. If the photovoltaic output fluctuation component is greater than the load power fluctuation component, the operating scenario category is determined to be a combination of photovoltaic fluctuation and load stability. Otherwise, the operating scenario category is determined to be a combination of photovoltaic stability and random load.

[0027] Furthermore, the step of dynamically allocating weights to the two optimization objectives—minimizing the number of voltage exceedances and minimizing system bus loss—in the multi-objective optimization model according to the preset weight configuration corresponding to the determined operating scenario category includes:

[0028] For scenarios involving a combination of peak and off-peak photovoltaic loads, the weight allocated to minimizing system bus losses is higher than the weight allocated to minimizing the number of voltage overruns.

[0029] For scenarios involving a combination of photovoltaic off-peak and load peak, and a combination of photovoltaic fluctuation and load stability, the weight assigned to the objective of minimizing the number of voltage over-limit occurrences is higher than the weight assigned to the objective of minimizing system bus losses.

[0030] For scenarios involving a stable photovoltaic load and a random load combination, a balanced weight is assigned to the two optimization objectives.

[0031] Furthermore, the step of iteratively solving the multi-objective optimization model using the particle swarm optimization algorithm to obtain the optimized configuration parameters and operating strategy of the energy storage system includes:

[0032] In each iteration of the particle swarm optimization algorithm, a main optimization process including particle update and evaluation is executed, followed by a constraint repair process based on grid sensitivity information; wherein, the constraint repair process includes:

[0033] The system detects whether the candidate solutions corresponding to particles in the population violate the constraints on the safe operation of the power grid. If a particle violates the constraints, the system calculates the sensitivity matrix of the power grid safe operation indicators to the changes in node power injection based on the power grid topology data. The power grid safe operation indicators include at least transformer load rate and line loss.

[0034] Based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix, the decision variables of the particles that violate the constraints are directionally modified to generate modified particles that satisfy the constraints.

[0035] The corrected particles are reintroduced into the population and participate in subsequent iterative optimizations together with the original particles.

[0036] Furthermore, the power grid safety operation constraints include at least one of the following categories:

[0037] Anti-power supply constraint, which is configured to take effect in scenarios where the operating scenario category is a combination of photovoltaic peak and load off-peak, is used to limit the minimum charging power of the energy storage system in order to prevent excess photovoltaic power from being fed back to the upstream grid;

[0038] Transformer load rate constraint is used to limit the maximum load rate of transformers connected to energy storage in any given time period.

[0039] Line loss constraints are used to limit the maximum line loss in the distribution area caused by energy storage operation;

[0040] The power response speed constraint is configured to take effect in scenarios with a combination of fluctuating photovoltaic power and stable load, as well as scenarios with a combination of stable photovoltaic power and random load, to limit the minimum rate of change in energy storage power.

[0041] Furthermore, the step of directionally correcting the decision variables of the particles that violate the constraints based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix includes:

[0042] Based on the type of constraint violated, the corresponding preset repair logic is invoked. This repair logic, based on the sensitivity matrix, determines a correction strategy for the energy storage power value in the particle decision variable; wherein the correction strategy includes at least one of the following:

[0043] If the transformer load rate constraint is violated, the discharge power of nodes with high sensitivity to the transformer load rate will be reduced or their charging power will be increased in the corresponding time period, based on the sensitivity of node power injection to the transformer load rate. Energy compensation will be carried out in other time periods to maintain the charging and discharging balance.

[0044] If the line loss constraint is violated, the charging and discharging power of the nodes located upstream or downstream of the line will be adjusted according to the sensitivity of the node power injection to the specified line loss, so as to change the power flow distribution and reduce the line loss.

[0045] If the reverse power supply constraint is violated, the charging power of the node located electrical downstream of the photovoltaic central access point will be increased first, based on the sensitivity of the node power injection to the reverse power supply.

[0046] If the power response speed constraint is violated, the step changes in the power sequence that exceed the preset ramp rate limit will be smoothly decomposed and adjusted into continuous changes in multiple adjacent time periods.

[0047] In a second aspect, the present invention provides an electronic device, comprising: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, the instructions being executed by the one or more processors to cause the one or more processors to implement the method described above.

[0048] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0049] This invention achieves dynamic weight allocation of optimization objectives in a multi-objective optimization model by accurately determining the operational scenario category of the target distribution area. This ensures that the optimization objectives are precisely matched with the core operational risks under different scenarios in the distribution area, guaranteeing that the direction of energy storage optimization aligns with the actual operational needs of the distribution area. Simultaneously, during the iterative solution process of the particle swarm optimization algorithm, the traditional penalty-elimination constraint handling method is abandoned. Instead, grid sensitivity information calculated based on grid topology data is used to directionally correct candidate solutions that violate grid safety operation constraints. This effectively improves the search efficiency of the algorithm under complex constraints, avoids the algorithm getting trapped in local optima, ensures the feasibility of the iteratively obtained solutions, and maintains population diversity by retaining the corrected particles, thus helping the algorithm to obtain better results. Finally, based on the optimized configuration parameters and operating strategies obtained from the algorithm, the energy storage system in the distribution area is configured and controlled. This generates an energy storage control scheme adapted to the real-time and future preset operating states of the target distribution area, effectively balancing the safety and economy of distributed photovoltaic distribution area grid operation, adapting to the dynamic changes in photovoltaic output and load power, and improving the pertinence and effectiveness of the configuration and operation control of distributed photovoltaic distribution area energy storage systems. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0051] Figure 1 The flowchart of a method for generating a distributed photovoltaic (PV) area energy storage optimization strategy provided in the embodiments of this specification. Figure 1 ;

[0052] Figure 2 The flowchart of a method for generating a distributed photovoltaic (PV) area energy storage optimization strategy provided in the embodiments of this specification. Figure 2 ;

[0053] Figure 3 This is a block diagram of an electronic device provided in the embodiments of this specification. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0055] In related technologies, the penetration rate of distributed photovoltaic (PV) power in distribution network areas is continuously increasing. PV power generation is intermittent and fluctuating, while user load is also random. The combination of these two factors has a significant impact on the power quality and safe and stable operation of the distribution area. This is mainly manifested in the following ways: during periods of high PV generation and low load, reverse power supply problems can easily occur, causing voltage rises or even exceeding limits in the distribution area; during periods of drastic changes in sunlight or load, frequent voltage deviations and fluctuations in the distribution area can occur.

[0056] To address the aforementioned issues, related technologies can configure energy storage systems at the transformer substation level and schedule their charging and discharging behavior using optimization algorithms. These technologies often employ an energy storage optimization model based on an optimization algorithm (Particle Swarm Optimization, PSO) to determine the optimal scheduling strategy for energy storage by minimizing line losses and reducing voltage overruns. Such methods typically construct a multi-objective optimization model that includes constraints on the energy storage itself (such as state of charge (SOC) limits and charging / discharging power limits), and solve it using an improved multi-objective PSO algorithm. This model coordinates different optimization objectives (minimizing line losses and minimizing voltage overruns) based on fixed weights.

[0057] However, with the increasing complexity of application scenarios, this type of method has revealed the following systemic defects when facing the actual operation of distributed photovoltaic power stations, resulting in insufficient control accuracy, economy, and applicability. The root cause lies in the disconnect between the optimization logic and the actual physical operation law of the power grid:

[0058] Firstly, regarding constraint handling mechanisms, to meet practical engineering requirements (preventing transformer overload, line overheating, and backflow), related technologies need to introduce numerous complex power grid safety operation constraints into the optimization model. To address this, related technologies often employ a "penalty-elimination" mechanism, imposing a fixed penalty on candidate solutions that violate constraints or discarding them directly. In high-dimensional, nonlinear, and strongly constrained energy storage optimization scenarios, this passive approach leads to a significant consumption of computational resources on evaluating infeasible solutions, severely compressing the effective search space, making the algorithm prone to getting trapped in local optima, resulting in slow convergence speeds, and making it difficult to obtain high-quality feasible solutions that satisfy both strict engineering constraints and achieve multi-objective optimization within a limited number of iterations.

[0059] Secondly, in coordinating optimization objectives, related technologies often employ fixed weights to balance economic efficiency (reducing line losses) and safety (reducing voltage overruns). However, the operating scenarios of distribution transformer areas are dynamic: for example, during a surge in photovoltaic power at midday, the core challenge is suppressing reverse power supply, and economic efficiency should be prioritized; while during a surge in load during the evening peak, the core challenge is supporting voltage, and safety should be prioritized. Fixed-weight optimization strategies cannot adapt to these dynamically changing scenario priorities, resulting in potentially unsuitable optimization solutions and poor targeted adjustments.

[0060] In summary, in the context of energy storage configuration and operation optimization in distributed photovoltaic power stations, existing particle swarm optimization algorithms suffer from low search efficiency and inability to dynamically adapt optimization results to the power station operation scenario due to their "penalty-elimination" constraint processing mechanism and fixed-weight multi-objective configuration.

[0061] This embodiment provides a method for generating a distributed photovoltaic (PV) area energy storage optimization strategy. The execution subject of the method can be an edge computing device, a smart gateway, a cloud server, or an area energy storage collaborative controller, etc.

[0062] like Figure 1 and Figure 2 As shown, the method may include:

[0063] Step S12: Obtain real-time operating data, grid topology data, and photovoltaic output and load power prediction data for the target transformer area in the future preset time period.

[0064] In this embodiment, the target distribution area can be represented as a distribution network area including distributed photovoltaic units, electrical loads, distribution transformers, and energy storage systems. Specifically, it can be a single independent distribution area or a cluster of multiple adjacent and closely electrically connected distribution areas. The real-time operating data can be the dynamic operating parameters of various electrical equipment, photovoltaic systems, and loads in the target distribution area during its current operation. Its function is to reflect the current actual operating status of the distribution area and provide real-time data support for subsequent scenario determination.

[0065] In this embodiment, the power grid topology data can be a set of data that characterizes the electrical connection relationships between various electrical devices (distribution transformers, lines, meter boxes, photovoltaic access points, energy storage access points, etc.) within the target distribution area, as well as the electrical parameters of the devices themselves.

[0066] In this embodiment, the future preset time period can be a pre-set time period used for predicting photovoltaic power output and load power, and its duration can be flexibly set according to actual application needs. The photovoltaic power output prediction data can be data obtained by predicting the total output of all distributed photovoltaic units in the target area within the future preset time period based on historical data and a preset prediction model, used to predict the future trend of photovoltaic power output. The load power prediction data is data obtained by predicting the total power of all electrical loads in the target area within the future preset time period based on historical data and a prediction model, used to predict the future trend of load.

[0067] In one possible and specific implementation plan, data acquisition and preprocessing can be performed. Specifically, static data such as basic information of the transformer substation (voltage level, transformer capacity), meter box information (latitude and longitude, affiliation), and branch parameters (resistance, reactance) can be imported through a user interface table, with the system automatically performing format verification. Dynamic data such as voltage, current, photovoltaic output, load power, transformer load rate, and line loss of the transformer substation nodes can be acquired in real time through the API interface. Alternatively, a meteorological service platform can be connected to obtain 24-hour solar intensity forecast data (15-minute / time resolution), combined with historical 3-month photovoltaic output-solar intensity and load-time period correlation data, to output a 24-hour photovoltaic output and load power forecast curve through a time-series model. Kalman filtering can also be used to remove noise from the dynamic data, integrating the collected static, dynamic, and forecast data into a standardized dataset, stored according to "substation-date-time period" classification, providing data support for energy storage optimization.

[0068] In one possible and specific implementation plan, firstly, based on the imported meter box location data (such as latitude and longitude), the location and distribution of all meter boxes are automatically displayed on the map interface, and they are distinguished by the meter box identification number (such as the last four digits), realizing a single map visualization of the transformer substation assets. Then, relying on the domain knowledge of the operation and maintenance personnel, the topology connections between meter boxes are interactively constructed on the visualization interface according to the actual electrical connection relationship of the transformer substation, thereby accurately simulating the physical topology of the power grid. On this basis, key nodes can be further highlighted on the topology map, including transformer outlets, photovoltaic centralized access points, and peak load nodes.

[0069] Step S14: Based on real-time operation data and prediction data, determine the operation scenario category of the target transformer area in the current and future preset time periods, and dynamically assign weights to each optimization objective in the preset multi-objective optimization model based on the operation scenario category; wherein, the multi-objective optimization model aims to minimize the number of voltage over-limits and minimize the system bus loss, and its constraints include energy storage body operation constraints and grid safety operation constraints.

[0070] In this embodiment, the operating scenario category can be represented as a category characterizing the operating characteristics of a distribution area, based on its power balance state (the matching relationship between photovoltaic output and load power) and power fluctuation state (the degree of fluctuation in photovoltaic output or load power). Its function is to distinguish different operating states of the distribution area and provide a basis for optimizing the allocation of target weights. The multi-objective optimization model is used to solve the optimization configuration and operation strategy of the energy storage system. Specifically, the optimization objectives are minimizing the number of voltage overruns and minimizing system bus losses. These two objectives jointly determine the quality of the energy storage optimization configuration and operation strategy.

[0071] In this embodiment, the dynamic weight allocation can be expressed as flexibly adjusting the weight ratio of the two optimization objectives in the multi-objective optimization model according to the current and future preset time period operation scenario categories of the target transformer area, so that the weight configuration is adapted to the core operation risks under the scenario category.

[0072] In this embodiment, the energy storage system's operational constraints are constraints set on the operating state of the energy storage system based on its own performance characteristics and operational requirements. Their function is to ensure the safe, stable, and long-term operation of the energy storage system and to prevent damage to the energy storage system caused by overcharging, discharging, or frequent switching.

[0073] In this embodiment, the power grid safety operation constraint is a constraint condition set on the power grid operation status based on the safety operation requirements of the target distribution area power grid. Its function is to ensure the safe and stable operation of the distribution area power grid and avoid power grid faults (e.g., transformer overload, line overheating, reverse power supply, etc.).

[0074] In one possible and specific implementation scheme, the constraints include constraints on the operation of the energy storage unit and constraints on grid safety operation. Specifically, constraints on the operation of the energy storage unit may include constraints on the state of charge (SOC) of the energy storage unit, constraints on charge / discharge power, constraints on charge / discharge cycle life, and constraints on charge / discharge switching. The SOC constraint requires that the SOC of the energy storage unit be maintained within a preset range (e.g., 20% to 90%) to avoid damage to the energy storage battery due to overcharging or over-discharging. The charge / discharge power constraint requires that the absolute value of the charge / discharge power of the energy storage unit does not exceed its rated power to avoid exceeding the operating capacity of the energy storage unit. The charge / discharge cycle life constraint requires that the depth of charge / discharge in a single cycle does not exceed a preset threshold to extend the lifespan of the energy storage battery. The charge / discharge switching constraint aims to avoid frequent charge / discharge switching of the energy storage unit, with a switching interval of not less than a preset time (e.g., 30 seconds).

[0075] In one possible and specific implementation plan, the operating scenarios of the transformer substations can be divided into four categories based on real-time data and predicted data:

[0076] Scenario A (Peak PV + Low Load): PV output ≥ 120% of local load (11:00-14:00 noon), the core risk is reverse power supply.

[0077] Scenario B (PV off-peak + peak load): Load power ≥ 150% of PV output (7:00-9:00 AM and 6:00-8:00 PM), the core risk is low voltage.

[0078] Scenario C (PV fluctuations + stable load): PV output fluctuates by ≥20% within 1 minute, with voltage fluctuations being the core risk.

[0079] Scenario D (Stable PV + Random Load): Load power fluctuation within 1 minute is ≥15%, and the core risk is local voltage over-limit.

[0080] In this implementation scheme, dynamic weights can be assigned to the dual optimization objectives for each scenario:

[0081] The objective function can specifically be: ;in, Let F be the objective function of the multi-objective optimization model, which means minimizing the comprehensive optimization objective value F. F is a comprehensive index obtained by weighting the two sub-objectives of voltage over-limit number and system bus loss. It is a weighting coefficient that minimizes the number of times the voltage exceeds the limit. This indicates the total number of times the voltage of all grid nodes in the target area exceeds the allowable deviation range of the rated voltage within a preset time period. It is a weighting coefficient that minimizes system bus loss. This represents the total active power loss of all power distribution lines and transformers in the target area within a preset time period.

[0082] In this implementation scheme, the weighting rules may specifically be:

[0083] For scenario A, The purpose is to prioritize reducing line loss and suppressing reverse power supply.

[0084] For scenario B, The purpose is to prioritize reducing voltage over-limit.

[0085] For scenario C, The purpose is to prioritize stabilizing the voltage.

[0086] For scenario D, .

[0087] In one possible and specific implementation, a set of multi-dimensional, continuously quantified scenario evaluation metrics can be calculated in real time to achieve dynamic weighting. Core metrics may include:

[0088] The reverse power supply risk index is calculated by multiplying the ratio of surplus photovoltaic power (PV output minus load) to the rated capacity of the transformer by a distance correction factor related to the electrical topology of the distribution area. This distance correction factor can be calculated based on the maximum electrical distance between the PV injection point and the transformer outlet under the current power flow conditions, allowing the index to simultaneously reflect the degree of power surplus and its potential impact on the power grid.

[0089] Voltage deviation sensitivity can be calculated based on the node admittance matrix of the transformer area or through the real-time perturbation method. The voltage-active power sensitivity coefficient and voltage-reactive power sensitivity coefficient of the critical common coupling point (PCC) or the weakest voltage node are calculated and weighted to quantify the sensitivity of the node's voltage to power changes.

[0090] The power fluctuation intensity index is calculated by taking the standard deviation or coefficient of variation of photovoltaic power output and load power within a short time window, and then normalizing the two values ​​to take the larger value as a measure of the current intensity of system fluctuations.

[0091] Then, a built-in fuzzy inference engine can be implemented, taking as input the continuous indices (reverse power supply risk index, power fluctuation intensity index, and power fluctuation intensity index) calculated above. The engine's rule base is generated through training with expert experience or historical data, including a series of fuzzy rules such as "If the reverse power supply risk index is 'high' and the voltage deviation sensitivity is 'low,' then the voltage exceedance weight should tend to be 'lower.'" Through fuzzification, rule evaluation, and defuzzification processes, the engine directly outputs a set of continuous, interrelated dynamic weight values.

[0092] The purpose of this implementation scheme is to achieve a nonlinear, adaptive mapping from a multidimensional operating state space to an optimized weight space, so that the weights can respond smoothly and reasonably to subtle changes in the system state.

[0093] In a specific implementation plan, the reverse power supply risk index can be:

[0094]

[0095] In the formula, The reverse power supply risk index is the specific quantitative value of the first indicator. This is a function that takes the minimum value. The photovoltaic output value of the target transformer area during time period t. The load power value of the target transformer area during time period t. This represents the surplus photovoltaic power. A positive value indicates that the photovoltaic output is greater than the load power, indicating a power surplus. A negative value indicates that the photovoltaic output is less than the load power, indicating no power surplus. The rated apparent capacity of the distribution transformer in the target area. This is the rated power factor of the distribution transformer. This refers to the rated active power capacity of the transformer. Here, is the distance correction factor, , The maximum electrical distance between the photovoltaic injection point and the transformer outlet can be calculated based on the meter box location information and branch connection data in the power grid topology data, representing the farthest distance that excess photovoltaic power is transmitted to the transformer outlet. The total length of the line in the target transformer area.

[0096] Voltage deviation sensitivity can be:

[0097]

[0098] In the formula, Voltage deviation sensitivity is the specific quantitative value of the second indicator. This is the weighting factor for the voltage-active power sensitivity coefficient. The voltage-active power sensitivity coefficient characterizes the magnitude of voltage change at a selected node when the active power injection amount changes by a unit. It represents the absolute value of the voltage-active power sensitivity coefficient. This is the weighting factor for the voltage-reactive power sensitivity coefficient. It is the absolute value of the voltage-reactive power sensitivity coefficient, which is used to characterize the change in voltage at a selected node when the reactive power injection amount changes by a unit.

[0099] The power fluctuation intensity index can be:

[0100]

[0101] In the formula, The coefficient of variation is used to characterize the relative fluctuation of power data. The standard deviation within a 1-minute window. This represents the average power. This is the power fluctuation intensity index. This is the function for finding the maximum value. The coefficient of variation of photovoltaic output in the target area during time period t is calculated based on the standard deviation and average power of photovoltaic output within a 1-minute time window, and characterizes the relative fluctuation of photovoltaic output. The maximum historical coefficient of variation for photovoltaic power output in similar distribution areas can be obtained by statistically analyzing long-term operating data of similar distributed photovoltaic distribution areas. This represents the normalized photovoltaic fluctuation value. The load power variation coefficient of the target transformer area in time period t can be calculated based on the standard deviation and average power of the load power within a 1-minute time window, and characterizes the relative fluctuation of the load power. This is the historical maximum coefficient of variation of load power in similar distribution areas, which can be obtained by statistically analyzing long-term operating data of similar distributed photovoltaic distribution areas.

[0102] Step S16: The multi-objective optimization model is iteratively solved using the particle swarm optimization algorithm to obtain the optimized configuration parameters and operation strategy of the energy storage system. In each iteration, the following steps are performed: If it is determined that the candidate solution corresponding to a particle in the population violates the power grid safety operation constraints, the corresponding power grid sensitivity information is calculated based on the power grid topology data, and the decision variables of the particle are directionally corrected based on the power adjustment direction indicated by the power grid sensitivity information to generate a corrected particle that meets the power grid safety operation constraints. The corrected particle is then reintroduced into the population to participate in subsequent iterative optimization.

[0103] In this embodiment, the optimized configuration parameters of the energy storage system are a set of parameters used to determine the hardware configuration of the energy storage system. The operating strategy of the energy storage system is a set of rules used to guide the operation of the energy storage system. Its function is to clarify the charging and discharging state and charging and discharging power of the energy storage system in different time periods and different operating scenarios, so as to achieve optimized operation of the energy storage system.

[0104] In one possible and specific implementation, the optimized configuration parameters may include energy storage access location parameters, energy storage capacity parameters, energy storage converter matching parameters, and energy storage battery pack series-parallel configuration parameters, etc. Specifically, the energy storage access location parameters further include the grid topology node number, physical installation point attributes, and electrical connection terminal attributes. The energy storage capacity parameters characterize the energy storage system's energy storage capacity, determining the system's ability to mitigate photovoltaic fluctuations and absorb photovoltaic surplus; specifically, they include the rated capacity, effective capacity, and charge / discharge depth threshold of the energy storage battery pack. The energy storage converter matching parameters may further include the rated power, rated voltage, current regulation range, and power factor regulation range of the energy storage converter. The energy storage battery pack series-parallel configuration parameters specifically include the number of battery cells connected in series, the number of cells connected in parallel, and the number of battery pack groups.

[0105] In one possible and specific implementation, the operation strategy may include time-segmented charging and discharging power strategy, scenario-based charging and discharging state strategy, real-time deviation power adjustment strategy, and multi-energy storage unit collaborative charging and discharging strategy, etc.

[0106] More specifically, the time-segmented charging and discharging power strategy can be the core strategy for the operation of an energy storage system. It matches the time division of a future preset time period, determining the specific charging and discharging power values ​​and power direction definitions for each sub-period within that preset time period. This serves as the direct basis for the energy storage system to execute charging and discharging operations. Each sub-period within the preset time period can be divided into 15 minutes, 30 minutes, or 1 hour, depending on the required prediction accuracy. Each sub-period corresponds to a unique charging and discharging power value. The power direction is defined by positive and negative values. A positive value indicates that the energy storage system is in a charging state, and the power value is the charging power magnitude; a negative value indicates that the energy storage system is in a discharging state, and the power value is the discharging power magnitude. The charging and discharging power values ​​for each sub-period are determined based on photovoltaic output prediction data, load power prediction data, and the optimization results of a multi-objective optimization model. For example, higher charging power values ​​are configured for sub-periods during peak photovoltaic periods and off-peak load periods, while discharge power values ​​are configured to match the load deficit for sub-periods during off-peak photovoltaic periods and peak load periods.

[0107] Scenario-based charging and discharging status strategies can be core strategies that are bound to the target transformer area's operating scenario category. They determine the charging and discharging status priority, charging and discharging power adjustment limit, and charging and discharging mode of the energy storage system under different operating scenario categories, adapting to the core operational risk prevention and control needs of the transformer area under different scenarios. In scenarios with a combination of peak and off-peak solar power, the charging and discharging priority is prioritized for charging. The upper limit for adjusting the charging power is set to the maximum charging power of the energy storage system. A "local consumption" charging mode is adopted to absorb as much surplus solar power as possible and avoid reverse power supply. In scenarios with a combination of off-peak solar power and peak load, the charging and discharging priority is prioritized for discharging. The upper limit for adjusting the discharging power is set to the maximum discharging power of the energy storage system. A "load supplement" discharging mode is adopted to supplement the load power deficit in the transformer area and avoid low voltage. In scenarios with a combination of fluctuating solar power and stable load, the charging and discharging priority is dynamic power adjustment. There is no fixed charging and discharging priority. A dynamic adjustment range for charging and discharging power is set. A "fluctuation smoothing" power adjustment mode is adopted, which quickly adjusts the charging and discharging power according to the real-time fluctuations in solar power output to suppress voltage fluctuations. In scenarios with a combination of stable solar power and random load, the charging and discharging priority is balanced charging and discharging. A balanced value for charging and discharging power is set. A "voltage-line loss dual control" mode is adopted to reduce the system's bus line loss while ensuring voltage stability.

[0108] In one possible and specific implementation, four types of scenario-based constraints can be added to the existing SOC limit, charge / discharge power limit, and charge / discharge balance constraint. Specifically, these four types of scenario-based constraints may include:

[0109] 1. Reverse power supply constraint (specific to scenario A): Energy storage charging power ≥ (PV output - local load - reverse power supply threshold), and the reverse power supply threshold is set to 10% of the transformer's rated capacity (not limited to 10%, just a specific example to avoid excess power flowing into the upstream grid).

[0110] 2. Transformer load factor constraint (all scenarios): After the energy storage charging and discharging power is superimposed, the transformer load factor shall be ≤85% (to avoid overload damage to the equipment; the specific threshold parameter is not limited).

[0111] 3. Line loss constraint (all scenarios): Line loss caused by energy storage charging and discharging ≤ 90% of the upper limit of line design loss (to prevent line overheating, specific threshold parameters are not limited).

[0112] 4. Response speed constraint (specific to scenario C / D): Energy storage charging and discharging power adjustment delay ≤ 1 second (quickly respond to fluctuations, specific threshold parameters are not limited).

[0113] In one possible and specific implementation scheme, the specific set of constraints could be:

[0114] ;

[0115] In the formula, For energy storage state of charge constraints, Let t be the state of charge of the target distribution area's energy storage system during time period t. Each sub-period represents a future preset time period (24 hours). Due to the power constraints of energy storage charging and discharging, Let be the charging and discharging power of the energy storage system during time period t. This represents the absolute value of the rated maximum charging and discharging power of the energy storage system. To constrain the daily charge and discharge energy balance. This is for reverse power supply constraint. Let be the charging power of the energy storage system during time period t. The total photovoltaic output of the target distribution area during time period t. Let be the total load power of the target transformer area in time period t. The rated apparent capacity of the distribution transformer in the target area. This is the reverse power supply threshold. For transformer load rate constraints. Let be the total active power of the distribution transformer during time period t. Line loss constraints. Let t be the total active power loss of all distribution lines in the target area during time period t. This is the upper limit of the design loss of the power distribution line. For power response speed constraints. This represents the power variation of the energy storage system in adjacent time periods. For time step. This represents the rate of change of energy storage power (ramp rate).

[0116] In one possible and specific implementation, each particle can represent a complete energy storage optimization scheme, and the encoding structure can be: . Specifically, It can be represented as the energy storage access location, which can be selected from the key nodes marked on the topology map (1 or 2, to adapt to different transformer area sizes). It can be expressed as energy storage capacity. To centralize energy storage capacity, For edge energy storage capacity. This is represented as a 24-hour charge / discharge power matrix, where... For centralized energy storage power during time period t, The power of edge energy storage during time period t. Let be the synergy coefficient, where It is used to allocate the charging and discharging power ratio. Specifically, for scenario A, It could be 0.7. It could be 0.3, meaning centralized energy storage would dominate charging. Scenario B could be... It is 0.3. The value is 0.7. This means that edge energy storage dominates the discharge. For scenario C / D, and Both are 0.5, achieving a balanced response.

[0117] In one possible and specific implementation, after each generation of particle position updates, the P algorithm immediately initiates a constraint satisfaction check, synchronized with the algorithm's iteration rhythm. For four types of scenario-based constraints, the satisfaction of each particle (corresponding to a set of energy storage configurations and operation schemes) is checked one by one.

[0118] Specifically, if a particle violates any constraint, it is marked as a "particle to be repaired," and the "constraint deviation degree" is calculated to quantify the severity of the violation:

[0119] Transformer load rate constraint deviation: ( >0 indicates a violation, with larger values ​​indicating more severe violations); where, This refers to the transformer load rate constraint deviation. >0 indicates that the energy storage scheme corresponding to the particle causes the transformer load rate to exceed the safe limit of 85%.

[0120] Reverse power supply constraint deviation: ( >0 indicates a violation); where, This refers to the reverse power supply constraint deviation. A value greater than 0 indicates that the energy storage charging power corresponding to the particle is insufficient to absorb the remaining photovoltaic power, and there is a risk of reverse power supply exceeding the threshold. The larger the value, the greater the reverse power supply. Let be the charging power of the energy storage system during time period t. This represents the actual amount of violation of the anti-power supply constraint.

[0121] Line loss constraint deviation: ( >0 indicates a violation); where, This refers to the deviation of the line loss constraint. 0.9 represents the actual upper limit of the line loss constraint, and 0.9 is the safety margin coefficient.

[0122] Response speed constraint deviation: ( >0 indicates a violation. Among them, It is the response speed constraint deviation. It is the absolute value of the rate of change (ramp rate) of energy storage power. It is the maximum allowable rate of change of energy storage power (upper limit of ramp rate).

[0123] Specifically, when A value of 0-0.2 indicates a minor violation, at which point the basic repair operator can be activated. When... A value greater than 0.2 indicates a severe violation, allowing the activation of a deep repair operator to ensure targeted repair. In other words, real-time monitoring can be established. Once a particle is detected violating any constraint, it is marked as a "particle to be repaired," and the type and severity of the violated constraint are recorded, rather than directly imposing a penalty or discarding it.

[0124] This embodiment achieves dynamic weight allocation of optimization objectives in a multi-objective optimization model by accurately determining the target transformer area's operating scenario category. This ensures that the optimization objectives are precisely matched with the core operational risks under different scenarios in the transformer area, guaranteeing that the direction of energy storage optimization aligns with the actual operational needs of the transformer area. Simultaneously, during the iterative solution process of the particle swarm optimization algorithm, the traditional penalty-elimination constraint handling method is abandoned. Instead, grid sensitivity information calculated based on grid topology data is used to directionally correct candidate solutions that violate grid safety operation constraints. This effectively improves the algorithm's search efficiency under complex constraints, avoids the algorithm getting trapped in local optima, ensures the feasibility of the iteratively obtained solutions, and maintains population diversity by retaining corrected particles, thus helping the algorithm to obtain better results. Finally, based on the optimized configuration parameters and operating strategies obtained from the algorithm, the energy storage system in the transformer area is configured and controlled. This generates an energy storage control scheme adapted to the real-time and future preset operating states of the target transformer area, effectively balancing the safety and economy of distributed photovoltaic transformer area grid operation, adapting to the dynamic changes in photovoltaic output and load power, and improving the pertinence and effectiveness of the configuration and operation control of the distributed photovoltaic transformer area energy storage system.

[0125] In some embodiments, the step of acquiring real-time operating data of the target transformer area, grid topology data, and photovoltaic power output and load power prediction data for a future preset period includes:

[0126] Step S122: Continuously acquire the voltage, current, real-time photovoltaic output, and real-time load power of each monitoring point in the target area through the data acquisition interface, as real-time operation data.

[0127] In this embodiment, the data acquisition interface is a module capable of communicating with on-site monitoring equipment, such as a communication module based on power line carrier, private wireless network, industrial Ethernet, or public wireless network (4G / 5G). The monitoring point is the location in the target distribution area's power grid where measuring devices are installed, such as the low-voltage side outlet of the transformer, the beginning of each major distribution branch, the grid connection point of the distributed photovoltaic system, and representative load concentration points.

[0128] Step S124: Obtain the power grid topology description data of the target transformer area. The power grid topology description data includes at least meter box location information, branch electrical parameters and transformer rated parameters. Based on the meter box location information and branch electrical parameters, determine the electrical connection relationship between nodes to form power grid topology structure data for calculation.

[0129] In this implementation, power grid topology description data can be obtained through pre-entry and / or real-time updates. First, existing data such as distribution network power marketing ledgers, power grid engineering design documents, and power equipment ledgers can be retrieved. Then, the meter box location information, branch electrical parameters, and transformer rated parameters for the target distribution area are extracted. Specifically, meter box location information may include the unique identifier of all user meter boxes within the distribution area, the unique identifier of centralized meter boxes, their physical installation address, associated distribution feeder and branch line information, and temporary grid node identifiers. Branch electrical parameters may include the unique identifier of each distribution line branch, line length, conductor type, resistance per unit length, reactance per unit length, and line cross-sectional area. Transformer rated parameters may include the unique identifier of the distribution transformer, rated capacity, rated primary voltage, rated secondary voltage, rated current, short-circuit impedance percentage, no-load loss, and load loss. If the target distribution area undergoes power grid structure modifications, addition or replacement of photovoltaic / load / energy storage equipment, or line relocation, the power grid topology description data can be updated in real-time through manual on-site collection and background entry to ensure that the data remains consistent with the actual power grid structure of the distribution area.

[0130] In this implementation, all grid nodes participating in electrical calculations within the target distribution area can be uniquely and standardizedly identified. Specifically, this includes high- and low-voltage side nodes of distribution transformers, photovoltaic unit grid-connected nodes, load consolidation nodes, meter box associated nodes, the beginning and end nodes of distribution line branches, and connection nodes of main / branch lines. Each node is assigned a fixed numerical or alphanumeric identifier, replacing the original temporary identifier. Simultaneously, a one-to-one correspondence is established between meter box location information, branch electrical parameters, transformer rated parameters, and standardized node identifiers, ensuring that all topology description data can be accurately mapped to the corresponding grid nodes. Based on the association between meter boxes and feeders / branch lines in the meter box location information, combined with the beginning and end node information of line branches in the branch electrical parameters, a topology analysis algorithm is used to determine the connection relationships of all standardized grid nodes. This identifies direct and indirect connections between nodes, eliminates invalid virtual connections, and forms a grid topology connection table with nodes as the core and line branches as the connecting links. Finally, the determined power grid topology connection table can be structurally integrated with the meter box location information, branch electrical parameters, and transformer rated parameters that incorporate standardized node identifiers to construct power grid topology data.

[0131] Step S126: Obtain historical operating data and future weather forecast data, process them through a pre-trained time series forecast model to obtain the initial photovoltaic output and load power curves for the future preset period, and perform smoothing filtering on the initial photovoltaic output and load power curves to generate photovoltaic output and load power forecast data.

[0132] In this embodiment, the historical operating data can be time series data of photovoltaic output and load power recorded in the target area over the past few months or even years.

[0133] In this embodiment, the future weather forecast data can be obtained from professional meteorological service agencies and can include future weather forecast information for the target area, such as irradiance (watts / square meter), ambient temperature (degrees Celsius), and cloud cover.

[0134] In this embodiment, the pre-trained time-series prediction model can be a model pre-constructed using machine learning algorithms, employing historical operational data and corresponding historical meteorological data as training samples, capable of capturing the changing patterns of photovoltaic output or load power with time and meteorological conditions. It can employ single models such as Long Short-Term Memory Networks, Gated Recurrent Units, and Support Vector Machine Regression, or a combination of multiple models. Specifically, in use, future meteorological forecast data, along with features such as date and time type, can be input into the pre-trained model, which then outputs the initial photovoltaic output curve and initial load power curve (i.e., the unsmoothed sequence of original predicted values) for a preset future time period (e.g., the next 24 hours).

[0135] In this embodiment, the smoothing filtering process can specifically be a Kalman filter algorithm, or it can be a moving average or a low-pass filter.

[0136] In some implementations, the step of determining the operational scenario category of the target transformer area in the current and future preset time periods based on real-time operational data and predicted data, and dynamically assigning weights to each optimization objective in a preset multi-objective optimization model based on the operational scenario category, includes:

[0137] Step S142: Calculate at least three types of scenario characteristic indicators based on the real-time operating data, prediction data, and power grid topology data; wherein, the scenario characteristic indicators include: a first indicator for characterizing power surplus and reverse power supply risk, a second indicator for characterizing the sensitivity of node voltage to power changes, and a third indicator for characterizing the severity of photovoltaic output or load power fluctuations.

[0138] Step S144: Based on the values ​​of at least three types of scenario characteristic indicators and preset scenario determination rules, determine the current operating scenario category; wherein, the operating scenario category includes at least the scenario of photovoltaic peak and load trough combination, the scenario of photovoltaic trough and load peak combination, the scenario of photovoltaic fluctuation and load stability combination, and the scenario of photovoltaic stability and load random combination.

[0139] In this embodiment, the preset scene determination rule can be represented as a set of rules pre-stored in the execution entity and used to determine the category of the running scene. The rule can be based on the comparison results of three types of scene feature indicators with preset thresholds, combined with the determination of power balance state, fluctuation state, and fluctuation dominant source, so as to realize the mapping of scene categories.

[0140] Step S146: Based on the preset weight configuration corresponding to the determined operating scenario category, dynamically assign weights to the two optimization objectives in the multi-objective optimization model: minimizing the number of voltage overruns and minimizing system bus loss; wherein, the weight configuration under different operating scenario categories is adapted to the core operating risks that need to be prioritized and controlled under that category.

[0141] In this embodiment, the preset weight configuration can be represented as a set of weights pre-stored in the execution entity, corresponding one-to-one with the four types of running scenarios. Each weight configuration includes the weight percentage of two optimization objectives (minimizing the number of voltage overruns and minimizing system bus loss). The sum of the weight percentages of the two optimization objectives is 1, and the magnitude of the weight percentage represents the priority of the corresponding optimization objective.

[0142] In some implementations, the step of calculating at least three types of scenario characteristic indicators based on the real-time operating data, the predicted data, and the power grid topology data includes:

[0143] Step S1422: Calculate the remaining photovoltaic power based on the current and predicted photovoltaic output and load power, and calculate the first index by combining the rated active power capacity of the transformer and the distance correction factor determined based on the electrical distance between the photovoltaic injection point and the transformer outlet in the electrical topology.

[0144] In this embodiment, firstly, the executing entity can calculate the time-series data of the remaining photovoltaic power. The executing entity can retrieve the real-time photovoltaic output and load power for the current time period, as well as the predicted photovoltaic output and load power for each time segment within a preset future time period. For each time period, the photovoltaic output value is subtracted from the load power value to obtain the remaining photovoltaic power for that time period. A positive value indicates power surplus, and a negative value indicates power deficit. Secondly, the executing entity can obtain the rated active power capacity of the transformer. The executing entity can read the rated apparent capacity and rated power factor of the main distribution transformer in the target area from the grid topology data, and multiply them to obtain the rated active power capacity of the transformer. If there are multiple transformers operating in parallel in the area, the sum of the rated active power capacities of each transformer is taken. Next, based on the grid topology data, the executing entity can analyze the electrical connection relationship between the grid connection point (photovoltaic injection point) of each photovoltaic power generation unit and the low-voltage side outlet of the transformer. The electrical distance between the two points is evaluated by calculating parameters such as the line impedance. Specifically, the closer the electrical distance, the larger the factor value, indicating less obstruction on the power backhaul path and higher risk; conversely, the farther the electrical distance, the smaller the factor value. Finally, for each time period, the executing entity can substitute the calculated remaining photovoltaic power, transformer rated active capacity, and distance correction factor into a preset weighted calculation formula to obtain the first index value for that time period.

[0145] Step S1424: Based on the power grid topology data, calculate the sensitivity coefficient of the voltage of the selected node in the distribution area to active power injection and the sensitivity coefficient to reactive power injection, and combine the two sensitivity coefficients according to the preset ratio to obtain the second index.

[0146] In this embodiment, the selected node can be a node within the transformer substation that requires key assessment of voltage stability, specifically the low-voltage side outlet of the transformer, historically weak voltage points, centralized photovoltaic grid connection points, and important load connection points.

[0147] In this embodiment, the executing entity can construct a node admittance matrix for analysis using branch parameters and connection relationships in the power grid topology data. Based on this matrix, power flow calculation or sensitivity analysis methods can be used to calculate the voltage-active power sensitivity coefficient and voltage-reactive power sensitivity coefficient for each selected node. Then, the executing entity can perform a weighted summation of the two sensitivity coefficients for the same node according to a preset ratio (for example, considering that active power often has a greater direct impact on voltage, a weight of 0.7 can be assigned to active power sensitivity and a weight of 0.3 to reactive power sensitivity), to obtain the voltage sensitivity of that node. Finally, the executing entity can perform a comprehensive processing of the voltage sensitivity of all selected nodes to form a second index representing the voltage sensitivity characteristics of the entire distribution area. The comprehensive method can be to take the maximum value of the sensitivity of all nodes (reflecting the weakest link), or to perform a weighted average based on the importance of the nodes (such as whether they are critical access points).

[0148] Step S1426: Based on the standard deviation and average power of photovoltaic output and load power within the preset time window, calculate the coefficient of variation of the two respectively, and take the maximum value after normalization of the historical maximum coefficient of variation to obtain the third index.

[0149] In this embodiment, the executing entity can pre-set a preset time window (e.g., 1 minute or 5 minutes) for analyzing short-term fluctuations. For the current time period and future time periods, the executing entity can extract continuous photovoltaic power output data sequences and load power data sequences within the duration of this window.

[0150] In this embodiment, the executing entity can perform calculations on the data sequence within each time window. Specifically, for the photovoltaic (PV) output sequence, the executing entity can calculate its standard deviation and average power within that window, and then divide the standard deviation by the average power to obtain the PV output coefficient of variation. Similarly, the same calculation is performed on the load power sequence to obtain the load power coefficient of variation. Next, the executing entity can perform normalization processing. The executing entity can retrieve the historical maximum coefficient of variation of PV output and load power from the historical database in long-term statistics. Dividing the currently calculated PV output coefficient of variation by its historical maximum value yields the normalized PV fluctuation value. Similarly, the executing entity can divide the load power coefficient of variation by its historical maximum value to obtain the normalized load fluctuation value. After normalization, both values ​​are between 0 and 1, with the closer to 1 indicating that the fluctuation is closer to the historical extreme case. Finally, for each time window, the normalized PV fluctuation value and the normalized load fluctuation value are compared, and the larger of the two is taken as the third indicator for that period.

[0151] In some implementations, the step of determining the current operating scenario category based on the values ​​of at least three types of scenario feature indicators and preset scenario determination rules includes:

[0152] Step S1442: Compare the value of the first indicator with the first preset threshold, and compare the value of the third indicator with the second preset threshold.

[0153] In this embodiment, firstly, the executing entity can compare the current value of the first indicator (power surplus and reverse power supply risk indicator) calculated in step S142 with a pre-set and stored first preset threshold. Next, it can compare the current value of the third indicator (power fluctuation intensity indicator) with another pre-set and stored second preset threshold.

[0154] In this embodiment, the first preset threshold and the second preset threshold can be threshold values ​​set based on statistical analysis of historical operating data of the target transformer area and combined with power grid safe operation experience, and are used to divide different state intervals.

[0155] Step S1444: Based on the comparison result between the first indicator and the first preset threshold, determine whether the current power balance state of the transformer area is a power surplus state or a power deficit state.

[0156] In this embodiment, the power surplus state refers to a power balance state in which the photovoltaic output of the target area exceeds the load power during a certain period, resulting in surplus photovoltaic power and potentially leading to risks such as reverse power supply and excessive line losses. The power deficit state refers to a power supply-demand gap in the target area where the photovoltaic output is less than the load power during a certain period.

[0157] In this embodiment, if the value of the first indicator is greater than or equal to the first preset threshold, the transformer area is determined to be in a state of power surplus. If the value of the first indicator is less than the first preset threshold, the transformer area is determined to be in a state of power deficit.

[0158] Step S1446: Based on the comparison result between the third indicator and the second preset threshold, determine whether the current power fluctuation state of the transformer area is a high fluctuation state or a low fluctuation state.

[0159] In this embodiment, the high-fluctuation state can be defined as a power fluctuation state in which the photovoltaic output or load power of the target area reaches a preset limit within a certain period of time, with drastic power changes, and high requirements are placed on the power response speed and adjustment accuracy of the energy storage system. The low-fluctuation state is a power fluctuation state in which the photovoltaic output or load power of the target area is lower than the preset limit within a certain period of time, with gentle power changes, and the energy storage system can operate according to conventional charging and discharging strategies.

[0160] In this embodiment, if the value of the third indicator is greater than or equal to the second preset threshold, the distribution area is determined to be in a high-fluctuation state. If the value of the third indicator is less than the second preset threshold, the distribution area is determined to be in a low-fluctuation state. This determination result indicates whether the system is experiencing significant short-term power changes.

[0161] Step S1448: Based on the combination of power balance state and power fluctuation state, map to a preset operating scenario category; wherein, if it is determined to be a power surplus state and a low fluctuation state, then the operating scenario category is determined to be a scenario of photovoltaic peak and load trough combination; if it is determined to be a power deficit state and a low fluctuation state, then the operating scenario category is determined to be a scenario of photovoltaic trough and load peak combination; if it is determined to be a power fluctuation state and a high fluctuation state, then based on the numerical relationship between the photovoltaic output fluctuation component and the load power fluctuation component constituting the third indicator, distinguish the dominant source of fluctuation; wherein, if the photovoltaic output fluctuation component is greater than the load power fluctuation component, then the operating scenario category is determined to be a scenario of photovoltaic fluctuation and load stability combination; otherwise, the operating scenario category is determined to be a scenario of photovoltaic stability and load random combination.

[0162] In this implementation, the executing entity can first filter out all time periods where the power fluctuation state is low. For these time periods, direct scenario category mapping is performed according to scenario mapping rules. The mapping logic can be as follows: if the state combination of a certain time period is "power surplus state + low fluctuation state", then the operating scenario category of that time period is directly determined to be a scenario of photovoltaic peak and load trough combination. If the state combination of a certain time period is "power deficit state + low fluctuation state", then the operating scenario category of that time period is directly determined to be a scenario of photovoltaic trough and load peak combination. The determination results of this type of time period can be recorded in real time, and corresponding operating scenario category identifiers can be added.

[0163] In this embodiment, the execution subject screen can select all time periods where the power fluctuation state is high. For such time periods, a fluctuation dominant source identification operation is first performed. Specifically, the magnitude relationship between the photovoltaic power output fluctuation component and the load power fluctuation component of the time period can be compared. If the photovoltaic power output fluctuation component is greater than the load power fluctuation component, the fluctuation dominant source of the time period is determined to be photovoltaic. If the load power fluctuation component is greater than or equal to the photovoltaic power output fluctuation component, the fluctuation dominant source of the time period is determined to be load.

[0164] In this embodiment, for high-fluctuation periods where the dominant source of fluctuation has been identified, the executing entity can map the scenario category according to the scenario mapping rules. The mapping logic can be as follows: if the dominant source of fluctuation in a certain high-fluctuation period is photovoltaic, then the operating scenario category for that period is determined to be a scenario of photovoltaic fluctuation and load stability; if the dominant source of fluctuation in a certain high-fluctuation period is load, then the operating scenario category for that period is determined to be a scenario of photovoltaic stability and load random combination.

[0165] In some implementations, the step of dynamically allocating weights to the two optimization objectives—minimizing the number of voltage exceedances and minimizing system bus loss—in a multi-objective optimization model according to a preset weight configuration corresponding to the determined operating scenario category includes:

[0166] Step S1462: For scenarios involving a combination of peak and off-peak photovoltaic loads, the weight assigned to the objective of minimizing system bus losses is higher than the weight assigned to the objective of minimizing the number of voltage overruns.

[0167] In this embodiment, the executing entity can retrieve a dedicated weight configuration corresponding to the scenario of photovoltaic peak and load trough combination from the preset weight configuration set of the local data storage unit. The weight ratio of this configuration can be in the range of: the weight ratio of minimizing the system bus loss target is 0.6-0.8, and the weight ratio of minimizing the number of voltage over-limit targets is 0.2-0.4. The specific value can be finely adjusted according to the anti-power supply prevention and control requirements and line loss control standards of the target transformer area, and the sum of the two weight ratios is always 1.

[0168] Step S1464: For scenarios involving a combination of photovoltaic off-peak and load peak, and a combination of photovoltaic fluctuation and load stability, the weight assigned to the objective of minimizing the number of voltage overruns is higher than the weight assigned to the objective of minimizing system bus losses.

[0169] In this embodiment, the executing entity can retrieve the general weight configuration corresponding to the two scenarios from the preset weight configuration set of the local data storage unit. The weight ratio of the configuration can be in the range of: the weight ratio of the target of minimizing the number of voltage over-limits is 0.6-0.8, and the weight ratio of the target of minimizing the system bus loss is 0.2-0.4.

[0170] Step S1466: For the scenario of a stable photovoltaic load and a random load combination, assign balanced weights to the two optimization objectives.

[0171] In this embodiment, the weight of the target of minimizing the number of voltage overruns is 0.5, the weight of the target of minimizing system bus loss is 0.5, and the sum of the two weights is 1, thus achieving a completely balanced priority configuration.

[0172] In some implementations, the step of iteratively solving the multi-objective optimization model using a particle swarm optimization algorithm to obtain the optimized configuration parameters and operating strategy of the energy storage system includes:

[0173] In each iteration of the particle swarm optimization algorithm, a main optimization process including particle update and evaluation is executed, followed by a constraint repair process based on grid sensitivity information; wherein, the constraint repair process includes:

[0174] Step S162: Detect whether the candidate solutions corresponding to particles in the population violate the constraints on power grid safe operation; if a particle violates the constraints, calculate the sensitivity matrix of the power grid safe operation index to the node power injection change based on the power grid topology data, wherein the power grid safe operation index includes at least transformer load rate and line loss.

[0175] Step S164: Based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix, the decision variables of the particles that violate the constraints are oriented to generate corrected particles that satisfy the constraints.

[0176] Step S166: The corrected particles are reintroduced into the population to participate in subsequent iterative optimization together with the original particles.

[0177] In this embodiment, in each iteration of the particle swarm optimization algorithm, in addition to executing the main optimization process of particle update and evaluation, a constraint repair process based on grid sensitivity information (steps S162, S164, S166) is also embedded to solve the problem that candidate solutions violate grid safety operation constraints and cannot be implemented during the iteration process. At the same time, the algorithm iteration efficiency and solution quality are improved through targeted constraint repair, so that the final energy storage configuration parameters and operation strategies meet both the requirements of multi-objective optimization and the grid safety operation specifications.

[0178] In this embodiment, constraint violation can be defined as a situation where the corresponding candidate solution (energy storage configuration and operation strategy) of a particle exceeds a preset safety threshold when the grid safety operation index exceeds the preset safety threshold during operation in the target distribution area's power grid. Node power injection change refers to the change in active / reactive power input / output at the nodes of the target distribution area's power grid. The sensitivity matrix is ​​a two-dimensional matrix constructed with "node power injection change" as columns and "grid safety operation index change" as rows. Each element in the matrix represents the "change in the corresponding grid safety operation index" caused by a "unit node power injection change," characterizing both the magnitude of the impact and indicating the direction of adjustment (positive / negative values ​​correspond to an increase / decrease in the index). Line loss is the active power loss caused by line resistance during power transmission in the target distribution area's distribution lines.

[0179] In one possible and specific implementation, firstly, population initialization / inheritance is performed. Specifically, the initial population of each iteration includes effective particles retained from the previous iteration (satisfying constraints and having good fitness), and a small number of newly generated particles (to maintain population diversity). Then, the main optimization process (particle update and evaluation) is executed. Specifically, a preset particle swarm optimization formula can be called, combined with the aforementioned dynamically allocated optimization objective weights, to update the decision variables (position) and update trend (velocity) of each particle. Then, based on the two optimization objectives of the multi-objective optimization model (minimizing the number of voltage overruns and minimizing system bus losses) and their corresponding weights, the fitness value of each particle is calculated (the better the fitness value, the better the optimization effect of the candidate solution). While evaluating the fitness, it is initially detected whether the particles violate the power grid safety operation constraints, and particles that need constraint repair are marked. Next, for the marked particles that violate the constraints, targeted correction, verification, and population reintroduction are performed, which is the core of this implementation. Then, the repaired particles are integrated with the original effective particles, and the individual optimal solution and the global optimal solution are updated to provide direction for the next iteration. Finally, if the iteration termination condition is met (maximum number of iterations reached, optimal solution converges and stabilizes), the iteration stops and the optimal solution is output; otherwise, the next iteration begins.

[0180] In one possible and specific implementation, in each iteration, the executing entity can verify whether each candidate solution represented by each particle in the population satisfies the grid safety operation constraints defined in step S14. This verification involves direct calculation and comparison of the mathematical expressions of the constraints. For example, when verifying the reverse power supply constraint, it can calculate whether the total energy storage charging power represented by the particle's decision variable at time period t is greater than or equal to the difference between the remaining photovoltaic power at that time period and the preset reverse power supply threshold. When verifying the transformer load rate constraint, it can calculate whether the total transformer load after adding the energy storage charging and discharging power at any time period t exceeds a preset percentage (85%) of its rated capacity. When verifying the line loss constraint, it can calculate whether the line loss caused by the energy storage charging and discharging behavior exceeds a preset loss limit. For particles marked as violating constraints, the sensitivity matrix of node power injection changes to grid safety operation indicators can be calculated based on the grid topology data obtained in step S12.

[0181] In one possible and specific implementation, the particle that has been corrected in step S164 and meets all the constraints (i.e., the corrected particle) can be reintroduced into the current population. The corrected particle will coexist with its original, uncorrected version (i.e., the original particle) in the population and participate in subsequent iterative optimizations, including velocity updates, position updates, and fitness evaluations.

[0182] In some implementations, the power grid safety operation constraints include at least one of the following categories:

[0183] The reverse power supply constraint is configured to take effect in scenarios where the operating scenario category is a combination of peak photovoltaic load and off-peak load. It is used to limit the minimum charging power of the energy storage system to prevent excess photovoltaic power from being fed back to the upstream grid.

[0184] Transformer load rate constraint is used to limit the maximum load rate of transformers connected to energy storage in any given time period.

[0185] Line loss constraints are used to limit the maximum line loss in the distribution area caused by energy storage operation.

[0186] The power response speed constraint is configured to take effect in scenarios with a combination of fluctuating photovoltaic power and stable load, as well as scenarios with a combination of stable photovoltaic power and random load, to limit the minimum rate of change in energy storage power.

[0187] In one possible and specific implementation, after each generation of particle position updates, the PSO algorithm immediately initiates a constraint satisfaction check, synchronized with the algorithm's iteration rhythm. For four types of scenario-based constraints, the satisfaction of each particle (corresponding to a set of energy storage configurations and operating schemes) is checked one by one:

[0188] Reverse power supply constraint (specific to scenario A): ;in, Let be the charging power of the energy storage system during time period t. The total photovoltaic output of the target distribution area during time period t. Let be the total load power of the target transformer area in time period t. This is the reverse power supply threshold.

[0189] Transformer load factor constraints (all scenarios): ;in, Let be the total active power of the distribution transformer during time period t. This refers to the rated apparent capacity of the distribution transformer.

[0190] Line loss constraints (all scenarios): ;in, Let t be the total active power loss of all distribution lines in the target transformer area during time period t. This is the upper limit of the design loss of the power distribution line. This is the safety margin coefficient.

[0191] Response speed constraints (specific to scenario C / D): ;in, This represents the power variation of the energy storage system in adjacent time periods. This represents the rate of change of energy storage power. This is the minimum permissible rate of change. This refers to the rated maximum charge and discharge power of the energy storage system.

[0192] In some implementations, the step of directionally correcting the decision variables of the particle that violated the constraint based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix includes:

[0193] Based on the type of constraint violated, the corresponding preset repair logic is invoked. This repair logic, based on the sensitivity matrix, determines a correction strategy for the energy storage power value in the particle decision variable; wherein the correction strategy includes at least one of the following:

[0194] If the transformer load factor constraint is violated, the discharge power of nodes with high sensitivity to the transformer load factor will be reduced or their charging power will be increased in the corresponding time period, based on the sensitivity of node power injection to the transformer load factor. Energy compensation will be carried out in other time periods to maintain the charging and discharging balance.

[0195] In this embodiment, the preset repair logic can be a set of decision variable correction rules that are pre-stored in the execution subject and correspond one-to-one with the four types of power grid safety operation constraints. Each repair logic can be formulated based on the power grid sensitivity matrix to determine the priority correction node, energy storage power adjustment direction, and auxiliary correction requirements when the corresponding constraint is violated.

[0196] If the line loss constraint is violated, the charging and discharging power of the nodes located upstream or downstream of the line is adjusted according to the sensitivity of the node power injection to the specified line loss, so as to change the power flow distribution and reduce the line loss.

[0197] If the reverse power supply constraint is violated, the charging power of the node located electrical downstream of the photovoltaic central access point will be increased first, based on the sensitivity of the node power injection to the reverse power supply.

[0198] If the power response speed constraint is violated, the step changes in the power sequence that exceed the preset ramp rate limit will be smoothly decomposed and adjusted into continuous changes in multiple adjacent time periods.

[0199] In one possible and specific implementation, when a candidate solution corresponding to a particle is detected, causing the transformer load rate to exceed a preset safety limit during one or more time periods, the execution entity can activate the transformer overload repair operator. First, the specific time period (t) causing the transformer load rate to exceed the limit and the energy storage node that contributes the most to this exceedance are identified. Specifically, this can be identified by analyzing the power injection value of each energy storage node during that time period and its sensitivity coefficient to the transformer load rate. Then, based on the sensitivity matrix, for nodes with high sensitivity to the transformer load rate, their power changes have a greater impact on the load rate. Therefore, the repair strategy can prioritize reducing the discharge power (or increasing the charging power) of high-sensitivity nodes during the over-limit time period (t) to most effectively alleviate the transformer overload. To ensure the total charging and discharging energy of the energy storage system is balanced within a 24-hour cycle (meeting the energy storage system's operational constraints), the aforementioned power adjustment in time period t needs to be compensated for in other adjacent time periods. Compensation adjustments can be preferentially performed during time periods and at nodes with low sensitivity to the transformer load rate to minimize secondary disturbances to the transformer load.

[0200] In one possible and specific implementation, when a candidate solution corresponding to a particle is detected to cause the loss of one or more lines to exceed a preset upper limit, a line over-loss repair operator can be activated. First, the lines with excessive losses and the time periods in which the losses occur are identified. Then, based on the line loss sensitivity matrix, the power adjustment direction that can effectively reduce the loss of the line is determined. The repair strategy can change the power flow distribution through the line by adjusting the charging and discharging power of energy storage nodes located upstream or downstream of the line. For example, the discharging power of energy storage nodes located downstream of the line can be reduced to decrease power injection from the end of the line. Alternatively, the charging power of energy storage nodes located upstream of the line can be increased to increase power extraction from the beginning of the line.

[0201] In one possible and specific implementation, when a candidate solution corresponding to a particle fails to enable the total charging power of the energy storage system to absorb photovoltaic surplus and avoid reverse power supply under a peak photovoltaic load and off-peak load scenario (Scenario A), a reverse power supply constraint repair operator can be activated. This strategy is used to encourage the energy storage system to absorb photovoltaic surplus power at the most efficient location. According to the sensitivity matrix, nodes located electrically downstream of the photovoltaic centralized grid connection point have the most direct and significant effect on suppressing reverse power supply by increasing their charging power. Therefore, the repair strategy can prioritize increasing the charging power of energy storage nodes located electrically downstream of the photovoltaic centralized grid connection point during the corresponding time period.

[0202] In one possible and specific implementation, when, under a photovoltaic or load fluctuation scenario (Scenario C / D), the power change rate in adjacent time periods of the particle decision variable exceeds the maximum allowable ramp rate of the energy storage converter, a response speed repair operator can be activated. The power sequence characterizing the 24-hour charge / discharge plan can be scanned to identify adjacent time period pairs where the absolute value of the power change exceeds the maximum allowable rate of change. A local smoothing algorithm can be used to decompose the drastic step power change. Specifically, the power difference that originally occurred between time period t and t+1 and exceeded the ramp rate limit is distributed across multiple consecutive adjacent time periods (e.g., from t-1 to t+2). Through this decomposition, the discrete power step that did not meet the response speed constraint can be adjusted into a continuous, smooth power change curve, ensuring that the rate of change in each time step meets the preset ramp rate limit.

[0203] According to an embodiment of the present invention, an electronic device is provided; please refer to... Figure 3 The electronic device in this embodiment may include one or more of the following components: a processor, a network interface, memory, non-volatile memory, and one or more application programs, wherein the one or more application programs may be stored in non-volatile memory and configured to be executed by one or more processors, and the one or more programs are configured to perform the methods as described in the foregoing method embodiments.

[0204] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for generating an optimization strategy for distributed photovoltaic (PV) transformer area energy storage, characterized in that, include: Acquire real-time operating data, power grid topology data, and photovoltaic output and load power forecast data for the target distribution area in the future preset time period; Based on real-time operation data and forecast data, the operation scenario category of the target transformer area in the current and future preset time periods is determined, and the weights of each optimization objective in the preset multi-objective optimization model are dynamically assigned based on the operation scenario category; wherein, the multi-objective optimization model aims to minimize the number of voltage over-limits and minimize the system bus loss, and its constraints include energy storage body operation constraints and grid safety operation constraints. The multi-objective optimization model is iteratively solved using a particle swarm optimization algorithm to obtain the optimized configuration parameters and operating strategy of the energy storage system. In each iteration, the following steps are performed: if a candidate solution corresponding to a particle in the population is determined to violate the grid safety operation constraints, the corresponding grid sensitivity information is calculated based on the grid topology data. Based on the power adjustment direction indicated by this grid sensitivity information, the decision variables of the particle are directionally corrected to generate a corrected particle that satisfies the grid safety operation constraints. The corrected particle is then reintroduced into the population to participate in subsequent iterative optimizations.

2. The method according to claim 1, characterized in that, The steps of acquiring real-time operating data of the target distribution area, power grid topology data, and photovoltaic power output and load power prediction data for a future preset period include: Through the data acquisition interface, the voltage, current, real-time photovoltaic output, and real-time load power of each monitoring point in the target area are continuously acquired as real-time operation data. Obtain the power grid topology description data of the target transformer area. The power grid topology description data includes at least meter box location information, branch electrical parameters and transformer rated parameters. Based on the meter box location information and branch electrical parameters, determine the electrical connection relationship between nodes to form power grid topology structure data for calculation. Historical operational data and future weather forecast data are acquired and processed through a pre-trained time series prediction model to obtain the initial photovoltaic output and load power curves for a preset future period. The initial photovoltaic output and load power curves are then smoothed and filtered to generate photovoltaic output and load power prediction data.

3. The method according to claim 1 or 2, characterized in that, The step of determining the operational scenario category of the target transformer area in the current and future preset time periods based on real-time operational data and predicted data, and dynamically assigning weights to each optimization objective in the preset multi-objective optimization model based on the operational scenario category, includes: Based on the real-time operating data, forecast data, and power grid topology data, at least three types of scenario characteristic indicators are calculated; among them, the scenario characteristic indicators include: a first indicator for characterizing power surplus and reverse power supply risk, a second indicator for characterizing the sensitivity of node voltage to power changes, and a third indicator for characterizing the severity of photovoltaic output or load power fluctuations. Based on the values ​​of at least three types of scenario characteristic indicators and preset scenario determination rules, the current operating scenario category is determined; wherein, the operating scenario category includes at least the scenario of a combination of photovoltaic peak and load trough, a scenario of a combination of photovoltaic trough and load peak, a scenario of a combination of photovoltaic fluctuation and load stability, and a scenario of a combination of photovoltaic stability and random load. Based on the preset weight configuration corresponding to the determined operating scenario category, weights are dynamically assigned to the two optimization objectives in the multi-objective optimization model: minimizing the number of voltage overruns and minimizing system bus losses. The weight configuration under different operating scenario categories is adapted to the core operating risks that need to be prioritized and controlled under that category.

4. The method according to claim 3, characterized in that, The step of calculating at least three types of scenario characteristic indicators based on the real-time operating data, predicted data, and power grid topology data includes: The remaining photovoltaic power is calculated based on the current and predicted photovoltaic output and load power. The first index is calculated by combining the rated active power capacity of the transformer and the distance correction factor determined based on the electrical distance between the photovoltaic injection point and the transformer outlet in the electrical topology. Based on the power grid topology data, the sensitivity coefficients of the voltage of selected nodes in the distribution area to active power injection and reactive power injection are calculated, and the two sensitivity coefficients are combined according to a preset ratio to obtain the second index. Based on the standard deviation and average power of photovoltaic output and load power within a preset time window, the coefficient of variation of the two are calculated respectively. After normalization by the historical maximum coefficient of variation, the maximum value is taken to obtain the third index.

5. The method according to claim 4, characterized in that, The step of determining the current operating scenario category based on the values ​​of at least three types of scenario feature indicators and preset scenario determination rules includes: The value of the first indicator is compared with the first preset threshold, and the value of the third indicator is compared with the second preset threshold. Based on the comparison results between the first indicator and the first preset threshold, it is determined that the current power balance state of the transformer area is either a power surplus state or a power deficit state. Based on the comparison results between the third indicator and the second preset threshold, it is determined whether the current power fluctuation state of the transformer area belongs to a high fluctuation state or a low fluctuation state. Based on the combination of power balance state and power fluctuation state, a preset operating scenario category is mapped. If the state is determined to be a power surplus state with low fluctuation, the operating scenario category is determined to be a combination of photovoltaic peak and load trough. If the state is determined to be a power deficit state with low fluctuation, the operating scenario category is determined to be a combination of photovoltaic trough and load peak. If the power fluctuation state is determined to be a high fluctuation state, the dominant source of fluctuation is distinguished based on the numerical relationship between the photovoltaic output fluctuation component and the load power fluctuation component, which constitute the third indicator. If the photovoltaic output fluctuation component is greater than the load power fluctuation component, the operating scenario category is determined to be a combination of photovoltaic fluctuation and load stability. Otherwise, the operating scenario category is determined to be a combination of photovoltaic stability and random load.

6. The method according to claim 5, characterized in that, The step of dynamically allocating weights to the two optimization objectives—minimizing the number of voltage exceedances and minimizing system bus loss—in the multi-objective optimization model according to the preset weight configuration corresponding to the determined operating scenario category includes: For scenarios involving a combination of peak and off-peak photovoltaic loads, the weight allocated to minimizing system bus losses is higher than the weight allocated to minimizing the number of voltage overruns. For scenarios involving a combination of photovoltaic off-peak and load peak, and a combination of photovoltaic fluctuation and load stability, the weight assigned to the objective of minimizing the number of voltage over-limit occurrences is higher than the weight assigned to the objective of minimizing system bus losses. For scenarios involving a stable photovoltaic load and a random load combination, a balanced weight is assigned to the two optimization objectives.

7. The method according to claim 6, characterized in that, The step of iteratively solving the multi-objective optimization model using the particle swarm optimization algorithm to obtain the optimized configuration parameters and operating strategy of the energy storage system includes: In each iteration of the particle swarm optimization algorithm, a main optimization process including particle update and evaluation is executed, followed by a constraint repair process based on grid sensitivity information; wherein, the constraint repair process includes: The system detects whether the candidate solutions corresponding to particles in the population violate the constraints on the safe operation of the power grid. If a particle violates the constraints, the system calculates the sensitivity matrix of the power grid safe operation indicators to the changes in node power injection based on the power grid topology data. The power grid safe operation indicators include at least transformer load rate and line loss. Based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix, the decision variables of the particles that violate the constraints are directionally modified to generate modified particles that satisfy the constraints. The corrected particles are reintroduced into the population and participate in subsequent iterative optimizations together with the original particles.

8. The method according to claim 7, characterized in that, The power grid safety operation constraints include at least one of the following categories: Anti-power supply constraint, which is configured to take effect in scenarios where the operating scenario category is a combination of photovoltaic peak and load off-peak, is used to limit the minimum charging power of the energy storage system in order to prevent excess photovoltaic power from being fed back to the upstream grid; Transformer load rate constraint is used to limit the maximum load rate of transformers connected to energy storage in any given time period. Line loss constraints are used to limit the maximum line loss in the distribution area caused by energy storage operation; The power response speed constraint is configured to take effect in scenarios with a combination of fluctuating photovoltaic power and stable load, as well as scenarios with a combination of stable photovoltaic power and random load, to limit the minimum rate of change in energy storage power.

9. The method according to claim 7 or 8, characterized in that, The step of directionally correcting the decision variables of particles that violate constraints based on the type of constraint violated and the adjustment direction indicated by the sensitivity matrix includes: Based on the type of constraint violated, the corresponding preset repair logic is invoked. This repair logic, based on the sensitivity matrix, determines a correction strategy for the energy storage power value in the particle decision variable; wherein the correction strategy includes at least one of the following: If the transformer load rate constraint is violated, the discharge power of nodes with high sensitivity to the transformer load rate will be reduced or their charging power will be increased in the corresponding time period, based on the sensitivity of node power injection to the transformer load rate. Energy compensation will be carried out in other time periods to maintain the charging and discharging balance. If the line loss constraint is violated, the charging and discharging power of the nodes located upstream or downstream of the line will be adjusted according to the sensitivity of the node power injection to the specified line loss, so as to change the power flow distribution and reduce the line loss. If the reverse power supply constraint is violated, the charging power of the node located electrical downstream of the photovoltaic central access point will be increased first, based on the sensitivity of the node power injection to the reverse power supply. If the power response speed constraint is violated, the step changes in the power sequence that exceed the preset ramp rate limit will be smoothly decomposed and adjusted into continuous changes in multiple adjacent time periods.

10. An electronic device, characterized in that, include: A memory, and one or more processors communicatively connected to the memory; The memory stores instructions that can be executed by the one or more processors to cause the one or more processors to implement the method as described in any one of claims 1 to 9.

Images