Power distribution network two-stage optimization method considering energy storage life and multi-time scale coordination

Abstract

The invention provides a power distribution network two-stage optimization method considering energy storage life and multi-time scale coordination, and the method comprises the steps: S1, collecting the historical load power and photovoltaic power data of a power distribution network, and employing a K-Shape clustering algorithm based on a dynamic time warping distance to generate a plurality of typical day scenes and the occurrence probability of the typical day scenes; s2, establishing a two-stage stochastic optimization model, solving the two-stage stochastic optimization model based on a typical day scene, and outputting a day-ahead scheduling plan in the next 24 hours; and S3, on a plan execution day, taking a day-ahead scheduling plan as a benchmark, performing intra-day rolling optimization based on a real-time state of the power distribution network and short-term prediction data in a future rolling time domain, and realizing coordination and closed-loop correction of the two stages. Through a progressive technical scheme of scene generation, random optimization and closed-loop rolling, on the premise of ensuring safe and stable operation of the system, the operation economy, the photovoltaic absorption capability and the energy storage service life are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution networks, and more particularly to a two-stage optimization method for power distribution networks that considers energy storage lifetime and multi-timescale coordination. Background Technology

[0002] With the acceleration of the global energy transition and the deepening of the "dual carbon" goal, distributed photovoltaic (PV) power, as a clean and renewable energy source, is seeing a continuous increase in its penetration rate in power distribution networks, which is of great significance for optimizing the energy structure and promoting sustainable development. However, the randomness and volatility of PV power output also pose serious challenges to the safe and stable operation and economic dispatch of power distribution networks.

[0003] Currently, there are still several technical problems that urgently need to be solved in the optimization and dispatching methods for distribution networks: First, the ability to cope with uncertainties is insufficient. Traditional day-ahead optimization methods rely heavily on a single deterministic prediction curve. Once the actual photovoltaic output and load deviate from the predicted values, the economic efficiency of the optimization scheme will decrease significantly, and even operational risks such as exceeding limits will be triggered. Often, costly real-time balancing measures are required for remediation. Second, the value of historical data is not fully utilized. Existing dispatching models mostly rely on pure physical modeling and do not effectively utilize the massive amount of operational data accumulated in the distribution network. This results in poor adaptability and robustness of optimization strategies, and the supporting role of data in dispatching decisions is not fully realized. Third, the lifespan management of key equipment such as battery energy storage systems is relatively crude. Existing dispatching strategies often ignore the cyclic degradation cost of energy storage or use overly simplified lifespan models, making it difficult to achieve economic operation of the system throughout its entire lifespan and affecting the long-term economic benefits and lifespan of energy storage. Fourth, the multi-timescale coordination mechanism is not flexible enough. The lack of effective constraints and feedback mechanisms between the day-ahead plan and intraday rolling optimization can easily lead to drastic fluctuations in intraday scheduling instructions. This may violate equipment operating rules and affect the execution of day-ahead market contracts, increasing the uncertainty and complexity of system operation and scheduling.

[0004] To address the aforementioned technical bottlenecks, there is an urgent need to propose a scheduling method that can effectively coordinate energy storage lifetime with multi-timescale optimization, so as to improve the economy, security and operational efficiency of the distribution network in a high-proportion photovoltaic environment. Summary of the Invention

[0005] To address the aforementioned technical problems, the first aspect of this invention proposes a two-stage optimization method for distribution networks that considers energy storage lifetime and coordination across multiple time scales.

[0006] The specific technical solution of the two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination in this invention is as follows: A two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination includes: Step S1: Collect historical load power and photovoltaic power data of the distribution network. After preprocessing and normalization, use the K-Shape clustering algorithm based on dynamic time warping distance to generate multiple typical daily scenarios and their occurrence probabilities. Step S2: Establish a two-stage stochastic optimization model with the objective of minimizing the total operating cost, which includes the interaction cost with the main grid, the cost of curtailment penalty, the cost of demand response compensation, and the cost of energy storage attenuation; based on the typical daily scenario, solve the two-stage stochastic optimization model and output the day-ahead scheduling plan for the next 24 hours. Step S3: On the planned execution day, based on the day-ahead scheduling plan, and based on the real-time status of the distribution network and short-term forecast data in the future rolling time domain, intraday rolling optimization is performed. In the intraday rolling optimization, anchoring constraints are introduced to limit the deviation between the optimization results and the day-ahead scheduling plan, thereby achieving coordination and closed-loop correction in two stages.

[0007] In some embodiments, step S1, which involves generating multiple typical daily scenes and their probabilities of occurrence using a K-Shape clustering algorithm based on dynamic time warping distance, includes: Use the elbow rule to determine the number of clusters; For the normalized dataset Perform clustering where each element is a daily curve. The objective function for clustering is: ; in, It is the first k Clusters, For the first k Cluster centers of each cluster For dynamic time-normalized distance; Calculate the probability of each cluster occurring: ; in, For clusters The number of daily curves included; Finally obtained k A collection of typical daytime scenes Each typical daily scenario All include cluster centers and probability of occurrence .

[0008] In some embodiments, in step S2, the constraints of the two-stage stochastic optimization model include: The power exchange constraints between the distribution network and the main grid are as follows: ; in, for t time kThe exchange power between the distribution network and the main grid in this scenario. To maximize power exchange between the distribution network and the main grid; The constraints on curtailment of photovoltaic power are as follows: ; ; in, for t time k Photovoltaic power generation in the scenario for t time k The amount of curtailed solar power in this scenario for t time k The maximum power output of photovoltaic power generation in this scenario.

[0009] SOC The state equations are as follows: ; in, 、 They are respectively t+1, t time k State of charge in the scenario 、 These are the charging and discharging efficiencies of battery energy storage, respectively. Energy storage for batteries t time k Charging power in various scenarios Energy storage for batteries t time k Discharge power in the scenario The rated capacity for battery energy storage; SOC The upper and lower limits are as follows: ; in, 、 These are the minimum and maximum values ​​of the state of charge, respectively; The charging and discharging power constraints are as follows: ; in, 、 These are the maximum power for charging and discharging the battery, respectively. The charge / discharge mutual exclusion constraint is as follows: ; in, It is a binary variable; SOC The boundary conditions are as follows: ; in, 、 These represent the initial state of charge (SOC) and the final state of charge (SOC) of battery energy storage, respectively. Incentive-based demand response constraints are as follows: ; ; in, for t time k Demand response power in the scenario This represents the largest adjustment amount for demand response within a single time period; For a complete regulatory cycle, This represents the maximum overall regulatory amount required for demand response within a complete regulatory cycle. The power balance constraints are as follows: ; in, for t time k Load power in the scenario.

[0010] In some embodiments, in step S2, the objective function of the two-stage stochastic optimization model is: ; in, for t time k Total operating cost in the scenario for t time k The interaction cost between the distribution network and the main grid in this scenario. for t time k The penalty for discarding light in the scene for t time k Demand response compensation costs in various scenarios for t time k Energy storage degradation costs in various scenarios; Among them Obtain it using the following formula: ; in, The initial investment cost of the energy storage system, Rated energy storage capacity (kWh). This refers to the rated number of cycles the battery performs at a specified average depth of discharge. The average depth of discharge.

[0011] In some embodiments, step S3 includes: Read the real-time status of the power distribution network, including the real-time state of charge of energy storage; Obtain short-term load forecasts and short-term photovoltaic forecasts over a rolling time span from the current moment to the future. Using the real-time state as the initial state, and based on the short-term load forecast and the short-term photovoltaic forecast, a deterministic optimization problem is solved in the rolling time domain, and the deterministic optimization problem includes the anchoring constraint. Execute the scheduling instruction for the current moment in the rolling optimization results, and repeat the above process over time.

[0012] In some embodiments, the objective function of the deterministic optimization problem is: ; in, for t time k Total operating cost in the scenario for t time k The interaction cost between the distribution network and the main grid in this scenario. for t time k The penalty for discarding light in the scene for t time k Demand response compensation costs in various scenarios for t time k Energy storage degradation costs in various scenarios For the current moment, The length of the rolling time domain; Among them Obtain it using the following formula: ; in, The initial investment cost of the energy storage system, For the rated capacity of energy storage, This refers to the rated number of cycles the battery performs at a specified average depth of discharge. The average depth of discharge.

[0013] In some embodiments, in step S3, the anchoring constraint is: ; ; in, This refers to the interaction power with the main network in the current scheduling plan. This refers to the demand response power in the day-ahead dispatch plan. The maximum allowable deviation for interaction with the mainnet. The maximum allowable deviation for demand response.

[0014] Compared with existing technologies, the two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination provided by this invention has the following significant advantages: 1. Significantly improved the adaptability and economy of the scheduling scheme to uncertainty.

[0015] This invention employs a K-Shape clustering algorithm based on dynamic time warping distance to generate multiple typical daily scenarios and their probabilities from massive historical data, quantifying the randomness, volatility, and other uncertainties of actual load into discrete scenarios with probabilities. Furthermore, by establishing a two-stage stochastic optimization model with the objective of minimizing total cost, a day-ahead scheduling plan that is feasible and cost-optimal under various possible scenarios is obtained. This overcomes the shortcomings of traditional single deterministic optimization, which is sensitive to prediction bias, and enhances the economy and robustness of distribution network operation.

[0016] 2. It has achieved refined modeling and optimization of the entire life cycle cost of energy storage systems.

[0017] This invention abandons the crude management model of traditional scheduling that ignores or simplifies energy storage lifespan degradation, and innovatively internalizes energy storage degradation costs as part of the operating costs. By introducing a degradation cost coefficient based on the whole life cycle theory and linking it to charging and discharging power, the optimization model can actively avoid scheduling strategies that accelerate battery aging while pursuing short-term operating economics. This achieves an optimal trade-off between short-term operating costs and the long-term lifespan of energy storage, significantly improving the long-term economic benefits and return on investment of the energy storage system.

[0018] 3. It effectively solves the problem of rigid connection in multi-timescale optimization and achieves smooth closed-loop correction.

[0019] In the intraday rolling optimization stage, this invention innovatively introduces anchoring constraints to limit the real-time optimization results to the allowable deviation range of the day-ahead plan. This ensures that the intraday rolling optimization can make real-time and accurate adjustments by taking advantage of the high accuracy of ultra-short-term forecasts, while remaining faithful to the framework of the day-ahead global optimization. This achieves a close and flexible coordination between the day-ahead and intraday stages, improving the stability and feasibility of the scheduling plan and the overall operating efficiency of the system.

[0020] 4. It fully leverages the value of historical data, driving the shift in scheduling strategies from model-driven to data and model fusion-driven.

[0021] This invention utilizes advanced clustering analysis methods to deeply mine the massive amounts of load and historical photovoltaic data accumulated in the power distribution network. It automatically extracts typical scenarios that represent different operational patterns. This makes optimizing scheduling strategies no longer a purely physical model calculation, but a more adaptive and representative intelligent decision based on historical operational patterns, greatly improving the engineering practicality and intelligence level of scheduling strategies.

[0022] In summary, this invention systematically addresses the core challenges of distribution network optimization and scheduling under high-proportion distributed photovoltaic (PV) access through a progressive technical solution involving scenario generation, random optimization, and closed-loop rolling. While ensuring the safe and stable operation of the system, it effectively improves operational economy, PV absorption capacity, and energy storage lifespan, providing strong technical support for building an efficient, intelligent, and reliable modern distribution network. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is an execution flowchart of the two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination in an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the distribution network optimization dispatching system in an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of the clustering results of a typical daily load scenario generated based on historical data in an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of the clustering results of a typical daily photovoltaic scenario generated based on historical data in an embodiment of the present invention.

[0028] Figure 5 This is a comparison chart of load forecast results for the day-ahead and intraday phases in an embodiment of the present invention.

[0029] Figure 6 This is a comparison chart of photovoltaic power prediction results for the day-ahead and intraday periods in an embodiment of the present invention.

[0030] Figure 7This is a comparison chart of net power predictions for the day-ahead and intraday periods in an embodiment of the present invention.

[0031] Figure 8 This is a comparison chart of the charging and discharging power plans for battery energy storage during the day-ahead and intraday periods in an embodiment of the present invention.

[0032] Figure 9 This is a comparison chart of the planned power interaction between the distribution network and the main grid in the day-ahead and intraday phases in an embodiment of the present invention.

[0033] Figure 10 This is a comparison chart of the invocation plans for demand response instructions in the day-ahead and intraday phases in an embodiment of the present invention. Detailed Implementation

[0034] To make the technical solutions and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail and completely below with reference to the accompanying drawings. The various embodiments described below are only some preferred embodiments of the present invention, and not all of them; the various embodiments described below are intended to explain the present invention and should not be construed as limiting the present invention; reasonable combinations of the technical features defined in the various embodiments of the present invention, as well as all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort, are all within the scope of protection of the present invention.

[0035] Example 1 like Figure 1 As shown, the present invention provides a technical solution: A two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination includes the following steps: Step S1: Collect historical load power and photovoltaic power data of the distribution network. After preprocessing and normalization, use the K-Shape clustering algorithm based on dynamic time warping distance to generate multiple typical daily scenarios and their occurrence probabilities.

[0036] The optional implementation process is as follows: The first step is to collect data.

[0037] Collect historical data for N consecutive days in the past (e.g., the past three years, approximately N=1095 days).

[0038] The data includes: load power for T time periods per day (e.g., 96 points, corresponding to 15-minute intervals). and photovoltaic power ,in, d For date indexing, 1 to N, t This is an intraday time-period index, from 1 to T.

[0039] Load power and photovoltaic power are the two most critical variables driving the operation of the distribution network. Choosing a 15-minute interval aligns with the common cycle of electricity market settlement and dispatch instructions, ensuring the model's engineering practicality.

[0040] Sufficiently long historical data (such as three years) can cover the operational status under different seasons (high load in summer and winter, low load in spring and autumn), different weather conditions (sunny, rainy, cloudy, and overcast), and different workday types (weekdays, weekends, and holidays), ensuring that the generated scenarios are comprehensive and representative.

[0041] The second step is to perform data preprocessing.

[0042] First, clean up abnormal data, that is, identify and remove obvious erroneous data points caused by sensor failure, communication interruption, etc. (such as negative power values, far exceeding the rated capacity, etc.).

[0043] Subsequently, missing data is filled in; that is, for the few missing data points, a global average interpolation method is used to fill in the gaps, and the formula is: ; in, Given a value M The number of known values. For supplementing missing values.

[0044] Finally, the data is normalized, that is, all daily curve data are normalized to... The interval, the formula is: ; in, , They are the first d The maximum values ​​of the overhead power and photovoltaic power.

[0045] Performing the above preprocessing on the data is key to ensuring the accuracy and reliability of subsequent cluster analysis results, and avoids interference from outliers and missing values ​​on the model.

[0046] Normalization is a standard preliminary step in cluster analysis. This is because the absolute value of load power can be very large (e.g., in the megawatt range), while photovoltaic power is relatively small (e.g., in the kilowatt range). Without normalization, the clustering algorithm would be dominated by variables with large absolute values. After normalization, the algorithm focuses more on the "shape" and "pattern of change" of each curve, rather than its absolute value.

[0047] The third step involves using a K-Shape clustering algorithm based on dynamic time warping distance to generate multiple typical daily scenes and their occurrence probabilities.

[0048] Optionally, the clustering process includes: First, the K-Shape clustering algorithm based on dynamic time regularization distance is selected as the clustering algorithm in this embodiment.

[0049] Subsequently, the elbow rule is used to determine the number of clusters K. This method calculates the sum of squared clustering errors corresponding to different K values ​​and selects the inflection point where the rate of error decrease suddenly slows down as the optimal K value, which can achieve a good balance between representational ability and model complexity.

[0050] Next, clustering is performed on the normalized dataset. Perform clustering, where each It is a daily curve, and the objective function for clustering is: ; in, It is the first k Clusters, It is the cluster center of this cluster (i.e., the typical daily curve). DTW ( ) is the dynamic time warped distance.

[0051] Finally, calculate the probability of each scenario and the probability of each scenario (cluster) occurring. It is determined by the proportion of the number of daily curves it contains to the total number, as follows: ; in, For clusters The number of daily curves included.

[0052] After the above clustering process, the final result is k A collection of typical scenarios Each scene From its central curve (The actual power value is obtained after inverse normalization) and its probability of occurrence constitute.

[0053] Traditional clustering algorithms (such as K-Means) typically use Euclidean distance. Euclidean distance requires the two sequences being compared to be strictly aligned, and it only cares about the numerical differences at the same point in time. This is very unfriendly to electricity load / PV curves because it cannot handle the "scaling" and "distortion" on the time axis.

[0054] Dynamic Time Warping (DTW) is a method for measuring the similarity between two time series of different lengths or out of sync. The core idea is to find the best alignment path between two series by allowing the series to be stretched or compressed non-linearly on the time axis to calculate the minimum cumulative distance between the two series.

[0055] The K-Shape algorithm is specifically designed for time series clustering. It uses a specific calculation method based on cross-correlation coefficients to update cluster centers, ensuring that the cluster centers retain the shape characteristics of the original time series. Simply put, K-Means' cluster centers are the arithmetic mean of all points, which may blur the peaks and troughs of the time series, while the cluster centers calculated by K-Shape are themselves a representative, clearly shaped, typical daily curve.

[0056] This embodiment employs the K-Shape clustering algorithm based on Dynamic Time Warping Distance (DTW) to perform clustering. DTW accurately measures the shape similarity between curves (solving the time misalignment problem), while K-Shape generates cluster centers that represent a class of curve shapes. The combination of these two methods enables highly accurate extraction of typical operating scenarios from historical data, such as "summer weekday peak curves," "spring weekend stable curves," and "cloudy day photovoltaic fluctuation curves."

[0057] Step S2: Establish a two-stage stochastic optimization model with the objective of minimizing the total operating cost, which includes the interaction cost with the main grid, the cost of curtailment penalty, the cost of demand response compensation, and the cost of energy storage attenuation. Based on the typical daily scenario, solve the two-stage stochastic optimization model and output the day-ahead scheduling plan for the next 24 hours.

[0058] As described in the background section, traditional optimization algorithms assume that tomorrow's photovoltaic power and load power will strictly follow the predicted curves, and formulate an "optimal" plan accordingly. Once the actual values ​​deviate from the predictions, this plan will experience performance degradation or even lead to safety issues.

[0059] The core idea of ​​stochastic programming optimization is to acknowledge the uncertainty of prediction: "Tomorrow may be one of K possible scenarios, each with a certain probability of occurrence. The goal is to develop a unique scheduling plan that minimizes the average cost (i.e., expected cost) across the K scenarios and is feasible in any of them."

[0060] The optional implementation process of step S2 is as follows: The first step is to perform mathematical modeling on each core device, which essentially involves generating constraints for a two-stage stochastic optimization model. These constraints include: 1. Power exchange constraints between the distribution network and the main grid, the formula is as follows: ; in, for t time k The exchange power with the main grid in this scenario (positive for electricity purchase, negative for electricity sales). To maximize power exchange with the main grid.

[0061] 2. The formula for constraining photovoltaic curtailment power is as follows: ; ; in, for t time k Photovoltaic power generation in the scenario for t time k The amount of curtailed solar power in this scenario for t time k The maximum power output of photovoltaic power generation in this scenario.

[0062] 3. SOC The state equations, specifically the formulas, are as follows: ; in, 、 They are respectively t+1, t time k State of charge in the scenario 、 These are the charging and discharging efficiencies of battery energy storage, respectively. Energy storage for batteries t time k Charging power in various scenarios Energy storage for batteries t time k Discharge power in the scenario The rated capacity for battery energy storage.

[0063] 4. SOC The upper and lower limits are defined using the following formulas: ; in, 、 These are the minimum and maximum values ​​of the state of charge, respectively.

[0064] 5. Charge and discharge power constraints, the specific formula is as follows: ; in, 、 These represent the maximum power for charging and discharging the battery, respectively.

[0065] 6. Charging and discharging mutual exclusion constraint, the specific formula is as follows: ; in, It is a binary variable.

[0066] 7. SOC Boundary conditions: ; in, 、 These represent the initial state of charge (SOC) and the final state of charge (SOC) of battery energy storage, respectively.

[0067] 8. Incentive-based demand response constraints, the formula is as follows: ; ; in, for t time k Demand response power in the scenario This represents the largest adjustment amount for demand response within a single time period. For a complete regulatory cycle, This refers to the maximum overall regulatory amount for demand response during a complete regulatory cycle.

[0068] In addition, the Big M method is used to solve the demand response correlation; if the user decides to respond to the demand, then... = 1. If the user does not respond to the request = 0, specifically as follows: ; ; in, Let be the power of user k's response at time t. It is a binary variable (0 or 1) that represents whether user k agrees to the response at time t (1 for response, 0 for no response). It is a very large positive number used to construct logical constraints. It is a very small positive number, representing the minimum power reduction threshold during response.

[0069] when When =0 (no response), the above formula is forced. =0.

[0070] when When =1 (response), the formula ensures Greater than a minimum value .

[0071] Constraints are imposed on the demand response period, including continuous response and intermittent time, as detailed below: ; ; in, This is the maximum limit for continuous response time. It represents the shortest interval between two demand responses.

[0072] 9. Power balance constraint, the formula is as follows: ; in, for t time k Load power in the scenario.

[0073] The second step is to determine the objective function of the two-stage stochastic optimization model as follows: ; in, for t time k Total operating cost in the scenario for t time k The interaction cost between the distribution network and the main grid in this scenario. for t time k The penalty for discarding light in the scene for t time k Demand response compensation costs in various scenarios for t time k Energy storage degradation costs in various scenarios; Among them Obtain it using the following formula: ; in, The initial investment cost of the energy storage system, Rated energy storage capacity (kWh). This refers to the rated number of cycles the battery performs at a specified average depth of discharge. The average depth of discharge.

[0074] The third step is to solve the two-stage stochastic optimization model described above and output the day-ahead scheduling plan for the next 24 hours.

[0075] Step S3: On the planned execution day, based on the day-ahead scheduling plan, and based on the real-time status of the distribution network and short-term forecast data in the future rolling time domain, intraday rolling optimization is performed. In the intraday rolling optimization, anchoring constraints are introduced to limit the deviation between the optimization results and the day-ahead scheduling plan, thereby achieving coordination and closed-loop correction in two stages.

[0076] The optional implementation process for step S3 is as follows: The first step is to read the real-time status of the distribution network, which should include at least the real-time state of charge of the energy storage. .

[0077] Step 2: Obtain information from the current time. Length of rolling time domain to the future Short-term load forecasts and short-term photovoltaic forecasts within the period.

[0078] The third step involves taking the real-time state as the initial state and using the short-term load forecast and short-term photovoltaic forecast as the basis to solve a deterministic optimization problem in the rolling time domain. This deterministic optimization problem includes anchoring constraints.

[0079] The objective function of this deterministic optimization problem is generally consistent with the day-ahead optimization in step S2, with the main difference being in the time range, as follows: ; in, for t time k Total operating cost in the scenario for t time k The interaction cost between the distribution network and the main grid in this scenario. for t time k The penalty for discarding light in the scene for t time k Demand response compensation costs in various scenarios for t time k Energy storage degradation costs in various scenarios For the current moment, The length of the rolling time domain.

[0080] The anchoring constraints are as follows: ; ; in, This refers to the interaction power with the main network in the current scheduling plan. This refers to the demand response power in the day-ahead dispatch plan. The maximum allowable deviation for interaction with the mainnet. The maximum allowable deviation for demand response.

[0081] Setting anchoring constraints ensures that intraday optimizations will not make drastic adjustments that contradict the overall optimization plan due to local forecast deviations, thus maintaining the stability and feasibility of the plan.

[0082] Execute the scheduling instructions for the current moment in the rolling optimization results, and repeat the above process over time until the planned execution day ends.

[0083] Example 2 This embodiment provides a two-stage optimization method for distribution networks that considers energy storage lifetime and multi-timescale coordination. The specific implementation method is the same as described above and will not be repeated here.

[0084] like Figure 2 As shown, the two-stage optimization method for distribution networks of the present invention is implemented in a modified IEEE 33-node distribution system. Photovoltaic power plants are connected at nodes 14, 17, and 20, with rated capacities of 600kW, 800kW, and 500kW, respectively. A battery energy storage system (BESS) with a rated capacity of 500kWh and a maximum charge / discharge power of 200kW is connected at node 13. The test is conducted using actual load and photovoltaic data (time resolution 15 minutes) from a certain location over a year.

[0085] The scene generated for step S1 is as follows: Figure 3 and Figure 4 As shown.

[0086] Time-of-use pricing uses the actual electricity price in that area, as shown in Table 1: Table 1. Time-of-use electricity pricing type Time period Price (RMB / kWh) peak 16:00-22:00 0.972 flat 07:00-15:00 0.669 valley 00:00-06:00,23:00-00:00 0.367 To fully verify the effectiveness of the two-stage optimization method for distribution networks of the present invention, the complete two-stage optimization method of the present invention is compared with the traditional optimization method based on a single day-ahead forecast, and the two-stage optimization method of the present invention without considering energy storage degradation, thereby verifying the effectiveness of the method provided by the present invention.

[0087] The day-ahead and intraday forecasts for load power, photovoltaic power generation, and net power are compared as follows: Figure 5 , Figure 6 , Figure 7 As shown in Table 2, the specific results of the operation are as follows.

[0088] Table 2. Comparison of running results type Total operating cost / yuan Traditional single optimization 29071.64 The method of this invention (without considering energy storage degradation) 28447.71 The method of this invention (considering energy storage degradation) 28616.36 By comparing the operational results of the three strategies, it can be found that the two-stage optimization method for distribution networks of the present invention has a significant advantage in reducing the total operating cost. The single optimization strategy, lacking real-time adjustment capability, has the highest cost, at 29,071.64 yuan. The two-stage optimization method for distribution networks of the present invention (without considering energy storage degradation), by introducing day-ahead and intraday two-stage optimization, effectively reduces the cost to 28,447.71 yuan, a reduction of 623.93 yuan compared to the single optimization. Furthermore, the two-stage optimization method for distribution networks of the present invention (considering energy storage degradation), while optimizing the allocation of power generation resources, also takes into account the lifespan of energy storage devices, resulting in a total operating cost of 28,616.36 yuan, a reduction of 455.28 yuan compared to the single optimization. This verifies the effectiveness of the present invention in balancing short-term operating costs and long-term equipment lifespan, making it suitable for power system optimization scenarios that require comprehensive consideration of economic benefits and equipment maintenance.

[0089] from Figure 8 , Figure 9 , Figure 10 As can be seen, this invention has significant advantages in power distribution network optimization. Firstly, the dynamic adjustment of energy storage plans (such as...) Figure 8 As shown, this allows energy storage devices to be flexibly charged and discharged at different times, reducing energy storage deviation and optimizing the power supply and demand balance. Secondly, real-time adjustment of grid interaction power (such as...) Figure 9 As shown, this significantly enhances system flexibility, enabling rapid response to load demand fluctuations and reducing the risk of supply-demand imbalances. Furthermore, demand response load adjustments (such as...) Figure 10 As shown, this invention effectively reduces load during peak hours, enhancing grid stability and operational efficiency. In summary, this invention not only improves the overall operational efficiency of the power system but also enhances flexibility and adaptability through real-time adjustments, providing strong support for the reliable operation of the power system.

[0090] In summary, the above verification analysis demonstrates that the two-stage optimization method for distribution networks proposed in this invention, which considers energy storage lifespan and multi-timescale coordination, significantly reduces operating costs through day-ahead and intraday optimization while also taking equipment lifespan into account. Dynamic adjustment of energy storage balances power supply and demand, real-time grid response enhances system flexibility, and demand response reduces peak load and optimizes operating efficiency, effectively balancing short-term operating costs and long-term equipment lifespan. This method is suitable for power system optimization scenarios that comprehensively consider economic benefits and equipment maintenance, providing an innovative and effective solution for distribution network optimization control.

[0091] Example 3 This invention provides a technical solution: An electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the two-stage optimization method for power distribution networks provided in the above embodiments.

[0092] Example 4 This invention provides a technical solution: A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the two-stage optimization method for power distribution networks provided in the above embodiments.

[0093] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, reasonable combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A two-stage optimization method for power distribution network considering energy storage life and multi-time scale coordination, characterized in that, It comprises: Step S1, collecting historical load power and photovoltaic power data of the power distribution network, after preprocessing and normalization, a plurality of typical day scenes and their occurrence probabilities are generated by using K-Shape clustering algorithm based on dynamic time warping distance; Step S2, a two-stage stochastic optimization model is established, the target is to minimize the total operation cost, the total operation cost includes the interaction cost with the main network, the light abandonment penalty cost, the demand response compensation cost and the energy storage attenuation cost; Based on the typical day scene, the two-stage stochastic optimization model is solved, and the day-ahead scheduling plan for the next 24 hours is output; Step S3, on the planning execution day, based on the real-time state of the power distribution network and the short-term prediction data in the future rolling time domain, the intra-day rolling optimization is carried out based on the day-ahead scheduling plan, in the intra-day rolling optimization, the anchor constraint is introduced to limit the deviation of the optimization result and the day-ahead scheduling plan, so as to realize the coordination and closed-loop correction of two stages.

2. The power distribution grid two-stage optimization method of claim 1, wherein, In step S1, the K-Shape clustering algorithm based on dynamic time warping distance is used to generate a plurality of typical day scenes and their occurrence probabilities, which comprises: The elbow rule is used to determine the number of clusters; On the normalized dataset Clustering is performed, where each element is a daily profile, the objective function of the clustering: ； wherein, is the k th cluster, is the k th cluster center, DTW is the dynamic time warping distance; The probability of occurrence of each cluster is calculated: ； wherein, is a cluster the number of diurnal curves contained in the cluster; A set of typical day scenarios is finally obtained k A set of typical day scenarios Each typical day scenario includes a cluster center and a probability of occurrence .

3. The power distribution grid two-stage optimization method of claim 1, wherein, In step S2, the constraint conditions of the two-stage stochastic optimization model include: The power exchange constraint between the power distribution network and the main network, specifically: ； wherein, is t moment k under the scenario of power exchange between the distribution network and the main network, is the maximum limit of power exchange between the distribution network and the main network; The photovoltaic light abandonment power constraint, specifically: ； ； Wherein, is t the moment k the power of photovoltaic power generation under the scene, is t the moment k the abandoned light power of photovoltaic power generation under the scene, is t the moment k the maximum power of photovoltaic power generation under the scene; SOC Equation of state, in particular: ； wherein, 、 respectively, t+1、t at the moment k under the scenario, 、 respectively, the charging and discharging efficiency of the battery energy storage, is the charging power of the battery energy storage under the scenario at the moment t at the moment k under the scenario, is the discharging power of the battery energy storage under the scenario at the moment t at the moment k under the scenario, is the rated capacity of the battery energy storage; SOC Upper and lower limits, in particular: ； wherein 、 SoCminand SoCmaxare minimum and maximum values of state of charge, respectively. The charging and discharging power constraint, specifically: ； wherein, 、 Pmax,chargeand Pmax,dischargeare the maximum power of the battery for charging and discharging, respectively. The charging and discharging mutual exclusion constraint, specifically: ； wherein is a binary variable; SOC Boundary conditions, in particular: ； wherein, 、 SoCini and SoCfin are the initial and final state of charge of the battery storage, respectively. The incentive demand response constraint, specifically: ； ； wherein, is t moment k demand response power in the scenario, is the maximum regulation amount of demand response in a single period; is a complete regulation period, is the maximum limit of the total regulation amount of demand response in the complete regulation period; The power balance constraint, specifically: ； wherein is t the moment k load power in the scenario.

4. The power distribution grid two-stage optimization method of claim 1, wherein, In step S2, the objective function of the two-stage stochastic optimization model is: ； Wherein, is t the moment k total operation cost in the scenario, is t the moment k interaction cost between the distribution network and the main network in the scenario, is t the moment k light abandonment penalty in the scenario, is t the moment k demand response compensation cost in the scenario, is t the moment k energy storage attenuation cost in the scenario; wherein is obtained by the formula: ； wherein, is the initial investment cost for the energy storage system, is the energy storage rated capacity, is the rated number of cycles for the battery at a specified average depth of discharge, is the average depth of discharge.

5. The power distribution grid two-stage optimization method of claim 1, wherein, Step S3 includes: Reading the real-time state of the power distribution network, the real-time state includes the real-time state of charge of the energy storage; Obtaining the short-term load prediction value and the short-term photovoltaic prediction value from the current time to the length of the rolling time domain; Taking the real-time state as the initial state, taking the short-term load prediction value and the short-term photovoltaic prediction value as the basis, a deterministic optimization problem is solved in the rolling time domain, the deterministic optimization problem contains the anchor constraint; The scheduling instruction of the current time in the rolling optimization result is executed, and the above process is repeated with time rolling.

6. The power distribution grid two-stage optimization method of claim 5, wherein, The objective function of the deterministic optimization problem is: ； wherein, is t the moment k the total operation cost in the scenario, is t the moment k the interaction cost between the distribution network and the main network in the scenario, is t the moment k the curtailment penalty in the scenario, is t the moment k the demand response compensation cost in the scenario, is t the moment k the energy storage attenuation cost in the scenario, is the current moment, is the rolling time domain length; wherein is obtained by the formula: ； wherein, is the initial investment cost for the energy storage system, is the energy storage rated capacity, is the rated number of cycles for the battery at a specified average depth of discharge, is the average depth of discharge.

7. The power distribution grid two-stage optimization method of claim 5, wherein, In step S3, the anchor constraint is: ； ； wherein, is an interaction power in the day-ahead dispatch plan to interact with the main grid, is a demand response power in the day-ahead dispatch plan to interact with the main grid, is a maximum deviation allowed to interact with the main grid, is a maximum deviation allowed to interact with the main grid.

Images