Multi-scenario-oriented capacity optimization method for grid-side hybrid energy storage system

Abstract

The application relates to the technical field of power grid configuration, and particularly discloses a power grid side hybrid energy storage system capacity optimization configuration method for multiple application scenarios, which comprises the following steps: collecting historical operation data of at least one monitoring point of the power grid side in a sampling period according to a fixed sampling time interval, and constructing a time sequence sample set after normalization processing and abnormal value elimination; and realizing differentiated response of the energy storage system to different power grid operation states through scene identification and classification, so that the traditional energy storage configuration method can avoid adopting a unified strategy, cannot take into account multiple demands such as frequency modulation, peak load shifting, fluctuation suppression and the like, and the situation of low energy storage resource utilization rate is avoided; and through dynamic adjustment and optimization of target weight, the energy storage system can play the best performance under different scenes, the adaptability and flexibility of the energy storage system are significantly improved, and the reliability of power grid operation after configuration is completed is improved.

Description

Technical Field

[0001] This invention relates to the field of power grid configuration technology, specifically to a method for optimizing the capacity configuration of a power grid-side hybrid energy storage system for multiple application scenarios. Background Technology

[0002] With the large-scale development of renewable energy power generation technologies, represented by wind power and photovoltaics, and the continuous increase in grid connection penetration, the operating characteristics of the power system have undergone profound changes. The intermittency, volatility, and uncertainty of renewable energy output have brought unprecedented challenges to the frequency stability, voltage support, power balance, and reserve capacity of the power grid.

[0003] Hybrid energy storage systems, by combining power-type and energy-type energy storage units, can simultaneously provide rapid power response and long-term energy support, becoming a key technical means for the grid to cope with the uncertainties of new energy sources. By combining the power level, energy capacity, response speed, and cycle life requirements of common scenarios such as peak shaving and valley filling, the capacity configuration of hybrid energy storage systems can be established to minimize the total life cycle cost of the system while improving the energy storage system's support capability for the grid.

[0004] In existing technologies, the capacity configuration of traditional hybrid energy storage systems is usually based on the power level, energy capacity, response speed, and cycle life requirements of common scenarios. However, in actual grid operation, scenarios such as frequency regulation, peak shaving and valley filling, fluctuation smoothing, and backup support occur alternately or even simultaneously. The requirements for the power level, energy capacity, response speed, and cycle life of energy storage systems vary significantly in different scenarios. Configuring based on fixed data will lead to insufficient frequency regulation response or poor fluctuation smoothing effect, thereby affecting the reliability of grid operation. Summary of the Invention

[0005] The purpose of this invention is to provide a method for optimizing the capacity configuration of grid-side hybrid energy storage systems for multiple application scenarios, and to solve the following technical problems: How to achieve scenario-adaptive capacity configuration for hybrid energy storage systems.

[0006] The objective of this invention can be achieved through the following technical solutions: A method for optimizing the capacity configuration of grid-side hybrid energy storage systems for multiple application scenarios, the method comprising: S1: Collect historical operating data from at least one monitoring point on the power grid side within a sampling period according to a fixed sampling time interval, and construct a time series sample set after normalization and outlier removal. The historical operating data includes active power, frequency deviation and voltage deviation. S2: Extract multi-dimensional feature vectors based on time series sample sets, and use clustering algorithms to divide the power grid operation status into multiple application scenarios, including frequency regulation, peak shaving and valley filling, fluctuation smoothing, and reserve support. Output the scenario label corresponding to each sampling time. ; S3: Establish a hybrid energy storage system topology model, including power-type energy storage units and energy-type energy storage units, and establish a comprehensive life cycle cost model, which includes purchase cost, replacement cost, operation and maintenance cost, loss cost and penalty cost; S4: Construct adaptive weight optimization models for different scenarios, and dynamically adjust the economic target weight, reliability target weight, and response speed target weight based on different scenario labels; S5: Establish a two-layer decoupled optimization model for power and energy. The upper-layer model uses the rated capacity of the energy-type energy storage unit as the decision variable to perform peak shaving and valley filling or reserve capacity matching. The lower-layer model uses the rated power of the power-type energy storage unit as the decision variable to perform high-frequency fluctuation suppression and frequency regulation command tracking. The upper and lower layers are iteratively coupled through the power imbalance correction factor. S6: The two-layer model described in S5 is solved using an optimization algorithm. Under the conditions of satisfying power balance constraints, state of charge constraints, and energy storage lifetime constraints, the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit are output. S7: By combining the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit output in S6, a hybrid energy storage system capacity configuration scheme is generated, which includes energy storage type, rated capacity, rated power, and installation location recommendations.

[0007] Furthermore, the multidimensional feature vector extracted in S2 includes: The volatility index, maximum power change rate, and energy gap duration feature vectors extracted from active power data in a time series sample set; Feature vectors of frequency over-limit amplitude and frequency over-limit duration extracted from frequency deviation data; Feature vectors of voltage over-limit amplitude and voltage over-limit duration extracted from voltage deviation data.

[0008] Furthermore, the process of dynamically adjusting the target weight in S4 includes: By combining the normalized demand intensity of economic, reliability, and response speed objectives under different scenario labels, the weights of these objectives are dynamically adjusted.

[0009] Furthermore, the process of establishing the power and energy two-layer decoupled optimization model in S5 includes: A first-order low-pass filtering method is used to allocate power between power-type energy storage units and energy-type energy storage units. The filtering time constant is adaptively determined by the scenario label. A second-level time constant is used in frequency modulation and fluctuation smoothing scenarios, while a minute-level time constant is used in peak shaving and valley filling scenarios and backup support scenarios, in order to distinguish the allocation paths of high-frequency power and low-frequency power.

[0010] Furthermore, the process of establishing the comprehensive lifecycle cost model in S3 includes: The total lifecycle cost is calculated by combining the initial purchase cost, replacement cost, operation and maintenance cost, life loss cost, and penalty cost.

[0011] Furthermore, the power imbalance correction factor in S5 includes: Through formula The power imbalance correction factor at time t is calculated; where t is any time. Let be the actual power demand of the power grid at time t. This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t.

[0012] Furthermore, the energy storage lifetime constraint in S6 includes: The cycle life of energy storage units satisfies the following relationship: ; Where T is the total running time. Let be the depth of discharge at time t. This represents the maximum number of cycles that the energy storage unit can withstand under the deep discharge conditions corresponding to time t. The cycle life of power-type energy storage units satisfies the following relationship: The cumulative value of the equivalent loop count shall not exceed the maximum allowed loop count.

[0013] Furthermore, the power balance constraint conditions in S6 include: The following relationship must be satisfied:

[0014] in, This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t. Let be the total power demand of the hybrid energy storage system at time t.

[0015] Furthermore, the state-of-charge constraints in S6 include: The state-of-charge constraints of energy storage units satisfy the following relationship: ; in, This is the lower limit threshold for the state of charge of the energy storage unit, with a value of 0.1. This is the upper limit threshold for the state of charge of the energy storage unit, with a value of 0.9. This represents the state of charge of the energy storage unit at time t. The state-of-charge constraints of power-type energy storage units satisfy the following relationship: ; in, This is the minimum threshold value for the state of charge of the power-type energy storage unit, taken as 0.05. This is the upper limit threshold for the state of charge of the power-type energy storage unit, with a value of 0.95. This represents the state of charge of a power-type energy storage unit at time t.

[0016] Furthermore, the process of generating the configuration scheme in S7 includes: S71: Based on the output results of the rated capacity of the optimal energy storage unit, match the standard specification energy storage products available on the market to determine the battery type, series and parallel connection method, and number of battery clusters of the energy storage unit; S72: Based on the output results of the optimal rated capacity of the power energy storage unit, match the standard specifications of the power energy storage products available on the market to determine the energy storage type and the number of modules of the power energy storage unit; S73: Calculate system-level configuration parameters, including total system rated power, total system energy capacity, power ratio and energy ratio of energy-type energy storage and power-type energy storage, and configure electrical interface parameters, including DC bus voltage, number of PCS units and power allocation, step-up transformer capacity, and access bay location; S74: Based on the comprehensive life cycle cost model, it outputs economic indicators such as initial investment, annual operation and maintenance cost, and levelized cost of energy storage. S75: Based on the monitoring point locations and power grid topology in S1, output installation location suggestions; S76: Generate a structured configuration scheme that includes the overall system configuration, details of energy storage units, electrical interface parameters, control strategies, economic indicators, and installation locations.

[0017] The beneficial effects of this invention are: (1) This invention realizes the differentiated response of the energy storage system to different grid operation states through scene recognition and classification, avoiding the situation where traditional energy storage configuration methods often adopt a unified strategy, which cannot take into account multiple needs such as frequency regulation, peak shaving and valley filling, and fluctuation smoothing, resulting in low utilization of energy storage resources. Furthermore, by dynamically adjusting and optimizing the target weight, the energy storage system can play its best performance in different scenarios, significantly improving the adaptability and flexibility of the energy storage system, thereby improving the reliability of grid operation after configuration.

[0018] (2) This invention adopts a power and energy decoupled optimization model and introduces a filtering time constant determined by scene label adaptively. In the capacity configuration stage, it solves the coupling problem of different time scale requirements in the same model. The upper model focuses on energy management and the lower model focuses on power response. The power imbalance correction factor realizes the synergistic coupling between the upper and lower layers. This dual-layer structure effectively avoids the problems of high optimization complexity and difficulty in convergence caused by power and energy coupling in the traditional single-layer model, and significantly improves the efficiency and stability of optimization solution.

[0019] (3) By combining a scenario-adaptive weight adjustment mechanism, this invention enables the optimization objectives of the energy storage system to be highly matched with the actual operating requirements. This avoids the situation where the fixed weight cannot adapt to the dynamic changes in the grid operation status, which often leads to poor performance of the energy storage system in certain scenarios. Therefore, the weight adjustment is driven by scenario labels. For example, in the frequency regulation scenario, the response speed weight is increased to ensure that the energy storage system can quickly track the frequency regulation command; in the peak shaving and valley filling scenario, the economic weight is increased to prioritize reducing the total life cycle cost of the energy storage system. This dynamic adaptation mechanism significantly improves the overall performance of the energy storage system.

[0020] (4) This invention decomposes the total life cycle cost into five components, covering the entire process of energy storage system from procurement, installation, operation, maintenance, replacement to decommissioning. Through the complete modeling of the five types of costs, the initial investment and long-term costs are automatically balanced during the optimization process. The output configuration scheme has the optimal total cost throughout the life cycle, avoiding the situation where traditional energy storage capacity configuration methods often only consider the initial purchase cost, but the total cost may be higher in actual operation due to frequent replacement, high maintenance costs, rapid aging, and penalty losses. This improves the optimization configuration of hybrid energy storage system capacity.

[0021] (5) This invention provides time-varying operating boundary conditions through scene recognition and dynamically adjusts the weight of the optimization target based on the boundary conditions, which solves the dynamic trade-off problem when the objectives conflict in multi-objective optimization. Finally, the time constant of power allocation is adaptively selected by scene label through a two-layer decoupling model. The three are not simply superimposed, but form an organic 'identification-decision-execution' closed-loop optimization logic, which produces a synergistic effect. It is this integrated design that enables the final capacity configuration to simultaneously meet the fast response of frequency regulation scenarios and the large capacity support of peak shaving and valley filling scenarios, thereby realizing the optimized configuration of the capacity of the hybrid energy storage system. Attached Figure Description

[0022] The invention will now be further described with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart of the capacity optimization configuration method for grid-side hybrid energy storage systems for multiple application scenarios in this invention; Figure 2 This is a flowchart of the process of generating a configuration scheme in S7 of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 As shown, in one embodiment, this application provides a method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios, the method comprising: S1: Collect historical operating data from at least one monitoring point on the power grid side within a sampling period according to a fixed sampling time interval, and construct a time series sample set after normalization and outlier removal. The historical operating data includes active power, frequency deviation and voltage deviation. S2: Extract multi-dimensional feature vectors based on time series sample sets, and use clustering algorithms to divide the power grid operation status into multiple application scenarios, including frequency regulation, peak shaving and valley filling, fluctuation smoothing, and reserve support. Output the scenario label corresponding to each sampling time. ; S3: Establish a hybrid energy storage system topology model, including power-type energy storage units and energy-type energy storage units, and establish a comprehensive life cycle cost model, which includes purchase cost, replacement cost, operation and maintenance cost, loss cost and penalty cost; S4: Construct adaptive weight optimization models for different scenarios, and dynamically adjust the economic target weight, reliability target weight, and response speed target weight based on different scenario labels; S5: Establish a two-layer decoupled optimization model for power and energy. The upper-layer model uses the rated capacity of the energy-type energy storage unit as the decision variable to perform peak shaving and valley filling or reserve capacity matching. The lower-layer model uses the rated power of the power-type energy storage unit as the decision variable to perform high-frequency fluctuation suppression and frequency regulation command tracking. The upper and lower layers are iteratively coupled through the power imbalance correction factor. S6: The two-layer model described in S5 is solved using an optimization algorithm. Under the conditions of satisfying power balance constraints, state of charge constraints, and energy storage lifetime constraints, the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit are output. S7: By combining the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit output in S6, a hybrid energy storage system capacity configuration scheme is generated, which includes energy storage type, rated capacity, rated power, and installation location recommendations. Through the above technical solution, this embodiment provides a method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios. First, historical operational data from at least one monitoring point on the grid side within a sampling period is collected at fixed sampling time intervals. After normalization and outlier removal, a time-series sample set is constructed. Then, multi-dimensional feature vectors are extracted from the time-series sample set, and a clustering algorithm is used to classify the grid operating state into multiple application scenarios, including frequency regulation, peak shaving and valley filling, fluctuation smoothing, and backup support. The scenario label corresponding to each sampling time is then output. A hybrid energy storage system topology model was established, including power-type energy storage units and energy-type energy storage units. A comprehensive life-cycle cost model was also established, including purchase cost, replacement cost, operation and maintenance cost, loss cost, and penalty cost. Subsequently, adaptive weight optimization models for different scenarios were constructed, and the economic target weights, reliability target weights, and response speed target weights were dynamically adjusted based on different scenario labels. A two-layer decoupled optimization model for power and energy was established. The upper-layer model uses the rated capacity of the energy-type energy storage unit as the decision variable to perform peak shaving and valley filling or reserve capacity matching, while the lower-layer model uses the rated capacity of the power-type energy storage unit as the decision variable. The rated power of the unit is the decision variable. High-frequency fluctuation suppression and frequency modulation command tracking are performed. The upper and lower layers are iteratively coupled through the power imbalance correction factor. Then, the S5 two-layer model is solved by the optimization algorithm. Under the conditions of satisfying the power balance constraint, state of charge constraint, and energy storage lifetime constraint, the optimal rated capacity of the energy type energy storage unit and the rated power of the power type energy storage unit are output. Combining the output optimal rated capacity of the energy type energy storage unit and the rated power of the power type energy storage unit, a hybrid energy storage system capacity configuration scheme including energy storage type, rated capacity, rated power, and installation location suggestions is generated. The working principle of the above technical solution is as follows: by collecting historical operating data (active power, frequency deviation, voltage deviation) from the power grid side, after normalization and outlier removal, a time series sample set is constructed. The power grid operating status is divided into four scenarios: frequency regulation, peak shaving and valley filling, fluctuation smoothing, and backup support using a clustering algorithm. Each moment is assigned a scenario label. Based on this, a hybrid energy storage system topology model and a full life cycle comprehensive cost model are established. A scenario-adaptive weight optimization model is constructed. The rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit are optimized respectively through a power and energy dual-layer decoupled optimization model. Finally, a configuration scheme including energy storage type, capacity, power, and installation location is output. By setting it up in this way, and through scene recognition and classification, the energy storage system can respond differently to different grid operating states. This avoids the situation where traditional energy storage configuration methods often adopt a uniform strategy, which cannot take into account multiple needs such as frequency regulation, peak shaving and valley filling, and fluctuation smoothing, resulting in low utilization of energy storage resources. Furthermore, by dynamically adjusting and optimizing the target weights, the energy storage system can perform at its best in different scenarios, significantly improving the adaptability and flexibility of the energy storage system, and thus improving the reliability of grid operation after configuration. Furthermore, by adopting a power and energy dual-layer decoupled optimization model, the division of labor between energy-type energy storage and power-type energy storage is clearly defined. The upper-layer model focuses on energy management, while the lower-layer model focuses on power response. The synergistic coupling between the upper and lower layers is achieved through a power imbalance correction factor. This dual-layer structure effectively avoids the problems of high optimization complexity and difficulty in convergence caused by power and energy coupling in traditional single-layer models, and significantly improves the efficiency and stability of optimization solutions. Based on this, time-varying operational boundary conditions are provided through scene recognition, and then the weights of the optimization objectives are dynamically adjusted based on these boundary conditions, solving the dynamic trade-off problem when objectives conflict in multi-objective optimization. Finally, a two-layer decoupling model is used to adaptively select the time constant of power allocation using scene labels, realizing the time scale decoupling of energy management and power control. The three are not simply superimposed, but form a closed-loop optimization logic of "identification-decision-execution". It is this integrated design that enables the final capacity configuration to simultaneously meet the fast response (second level) of frequency modulation scenarios and the large capacity support (hour level) of peak shaving and valley filling scenarios. This is a technical effect that cannot be achieved simultaneously by using single-scene optimization or fixed-structure optimization methods.

[0026] It should be noted that the optimization algorithm used in S6 is the multi-objective particle swarm optimization algorithm. The process in S2 of using a clustering algorithm to divide the power grid operation state into multiple application scenarios—frequency regulation, peak shaving and valley filling, fluctuation smoothing, and reserve support—includes that the clustering algorithm uses K-means clustering, where the number of clusters K=4, corresponding to the four scenarios of frequency regulation, peak shaving and valley filling, fluctuation smoothing, and reserve support. The distance metric used is Euclidean distance, calculated using the following formula: in and Let n be any two multidimensional feature vectors, and n be the total number of multidimensional feature vectors. In this invention, n is taken as 7, corresponding to the seven-dimensional features: volatility index, maximum power change rate, energy gap duration, frequency over-limit amplitude, frequency over-limit duration, voltage over-limit amplitude, and voltage over-limit duration. For the index of the feature dimension, For feature vectors In the Values ​​on the dimension, For feature vectors exist The values ​​on the dimension are selected, and the termination condition for algorithm iteration is set as a centroid change threshold less than 0.001 or reaching the maximum number of iterations of 100. The optimal number of clusters is determined by the elbow rule. That is, when K=4, the rate of decrease of the sum of squared errors within the cluster shows a clear inflection point.

[0027] The multidimensional feature vector extracted in S2 includes: The volatility index, maximum power change rate, and energy gap duration feature vectors extracted from active power data in a time series sample set; Feature vectors of frequency over-limit amplitude and frequency over-limit duration extracted from frequency deviation data; Feature vectors of voltage over-limit amplitude and voltage over-limit duration extracted from voltage deviation data; Through the above technical solution, this example provides a multi-dimensional feature vector extracted in S2, including a volatility index, a maximum power change rate, and an energy gap duration feature vector extracted based on active power data in the time series sample set. The method for extracting the volatility index is as follows: calculate the average active power at all sampling times, calculate the difference between the active power at each time and the average power, divide by the average power to obtain the relative power deviation, square the relative power deviation at each time, sum them, divide by the number of sampling points minus one, and finally take the square root. The result is the volatility index. The method for extracting the maximum power change rate is as follows: traverse all sampling times, calculate the absolute value of the active power difference between each adjacent sampling time, take the maximum value, and the result is the maximum power change rate. The method for extracting the duration of the energy gap is as follows: set the energy gap threshold to 80% of the average active power, identify all periods when the active power is continuously lower than the threshold, calculate the duration of each period, and take the maximum value among them. The result is the duration of the energy gap. This example also provides feature vectors for frequency over-limit amplitude and frequency over-limit duration extracted based on frequency deviation data. The extraction method for frequency over-limit amplitude and frequency over-limit duration is as follows: calculate the frequency deviation at each sampling time based on the rated frequency, set a primary frequency modulation dead zone threshold, mark the period when the absolute value of the frequency deviation continuously exceeds the threshold as a single frequency over-limit event, take the maximum value of the absolute value of the frequency deviation during the event as the frequency over-limit amplitude, and take the difference between the end time and the start time of the event as the frequency over-limit duration. This example also provides feature vectors for voltage over-limit amplitude and voltage over-limit duration extracted based on voltage deviation data. The extraction method for voltage over-limit amplitude and voltage over-limit duration is as follows: calculate the voltage deviation at each sampling time based on the rated voltage, set the upper and lower thresholds for qualified voltage, mark the period when the voltage value is continuously lower than the lower limit or continuously higher than the upper limit as a single voltage over-limit event, take the maximum absolute value of the voltage deviation during the event as the voltage over-limit amplitude, and take the difference between the end time and the start time of the event as the voltage over-limit duration. By setting it up in this way, feature vectors are extracted from three dimensions: active power, frequency deviation, and voltage deviation, a comprehensive characterization of the power grid's operating status is achieved. The active power feature reflects the energy supply and demand balance, the frequency deviation feature reflects the system inertia and frequency regulation requirements, and the voltage deviation feature reflects the reactive power and voltage support requirements. The features of the three dimensions are complementary, which enables the subsequent clustering algorithm to accurately distinguish different scenarios and avoids the scenario misjudgment problem caused by single-dimensional features.

[0028] The process of dynamically adjusting the target weight in S4 includes: By combining the normalized demand intensity of economic, reliability, and response speed objectives under different scenario labels, the weights of economic, reliability, and response speed objectives are dynamically adjusted. Through the above technical solution, this example provides a process for dynamically adjusting the weights of economic, reliability, and response speed objectives; Through formula The weights of the i-th optimization objective under scenario s are calculated, where s is the scenario label, taking values ​​of frequency modulation scenario, peak shaving and valley filling scenario, fluctuation smoothing scenario, and backup support scenario; i=1 represents the economic objective, corresponding to minimizing the total lifecycle cost; i=2 represents the reliability objective, corresponding to minimizing the power point tracking error; and i=3 represents the response speed objective, corresponding to minimizing the response time. This is the sensitivity adjustment factor, with a value range of [value range missing]. , Let be the normalized demand intensity of the i-th optimization objective in scenario s, with a value range of . By collecting no less than 500 sets of historical samples for each typical scenario, the normalized performance requirements of each optimization objective are statistically analyzed and the average value is taken; among them, the response speed requirement intensity is 0.9 in the frequency modulation scenario, the economic requirement intensity is 0.85 in the peak shaving and valley filling scenario, the reliability requirement intensity is 0.8 in the fluctuation smoothing scenario, and the reliability requirement intensity is 0.9 in the backup support scenario, where e is a natural constant. With this configuration, the scenario-adaptive weight adjustment mechanism ensures that the optimization goals of the energy storage system are highly matched with actual operational needs. This avoids the situation where fixed weights cannot adapt to dynamic changes in the grid's operating status, which often leads to poor performance of the energy storage system in certain scenarios. Therefore, by driving weight adjustment through scenario labels, for example, increasing the response speed weight in frequency regulation scenarios to ensure that the energy storage system can quickly track frequency regulation commands; and increasing the economic weight in peak shaving and valley filling scenarios to prioritize reducing the total life cycle cost of the energy storage system. This dynamic adaptation mechanism significantly improves the overall performance of the energy storage system. It should be noted that, The generation method involves collecting no fewer than 500 sets of historical samples for each typical scenario in any given scene, calculating the normalized performance requirements of each optimization objective, and then taking the average value. This method is for offline training; when applied online... For fixed lookup table values.

[0029] The process of establishing the power and energy two-layer decoupled optimization model in S5 includes: A first-order low-pass filtering method is used to allocate power between power-type energy storage units and energy-type energy storage units. The filtering time constant is adaptively determined by the scenario label. A second-level time constant is used in frequency modulation and fluctuation smoothing scenarios, while a minute-level time constant is used in peak shaving and valley filling scenarios and backup support scenarios, in order to distinguish the allocation paths of high-frequency power and low-frequency power. Through the above technical solution, this embodiment provides a process for establishing a two-layer decoupled optimization model of power and energy. Specifically, let the total power demand of the hybrid energy storage system be... The low-frequency component of the energy storage response is obtained after passing through a first-order low-pass filter. High-frequency components of power-type energy storage response in, The filtering time constants are as follows: 0.5 seconds for frequency modulation scenarios, 120 seconds for peak shaving and valley filling scenarios, 10 seconds for fluctuation smoothing scenarios, and 60 seconds for backup support scenarios. The Laplace operator is used; with this setting, the low-pass filtering method has the advantages of high computational efficiency and good real-time performance, making it suitable for online application scenarios of grid-side hybrid energy storage. By adaptively adjusting the time constant through scenario labels, the performance defects of a single time constant in different scenarios are avoided, thus maintaining the simplicity of the method while achieving adaptability to different scenarios.

[0030] The process of establishing the comprehensive lifecycle cost model in S3 includes: The comprehensive lifecycle cost is obtained by combining the initial purchase cost, replacement cost, operation and maintenance cost, life depreciation cost, and penalty cost. Specifically, it can be done through formulas Calculated; in, The initial purchase cost can be calculated using the formula. Calculated Cost per unit capacity For the rated capacity of energy storage, Cost per unit power Rated power of energy storage; To determine replacement costs, a formula can be used. Calculated Here, H is the discount rate, H is the total number of replacements, and h is the replacement frequency index. The lifespan of the energy storage unit; For operation and maintenance costs, the formula can be used. The calculation yields the result, where B represents the total lifespan in years, and b represents the year index. The maintenance cost for year b; The cost of lifespan depletion can be calculated using the formula... The calculation yields the result, where T is the total running time. Let be the depth of discharge at time t. This represents the maximum number of cycles that the energy storage unit can withstand under the deep discharge conditions corresponding to time t. Cost per unit of cycle loss; To incur penalties, a formula can be used. Calculated, where, The unit price for penalties for abandoning wind and solar power. For the amount of wind and solar power that has been curtailed, Penalty for power shortages per unit price Low battery; By setting it up in this way, the total lifecycle cost is broken down into five components, covering the entire process of energy storage systems from procurement, installation, operation, maintenance, replacement to decommissioning. Through the complete modeling of these five cost categories, the initial investment and long-term costs are automatically balanced during the optimization process. The output configuration scheme has the optimal total cost over the entire lifecycle, avoiding the situation where traditional energy storage capacity configuration methods often only consider the initial purchase cost, but the total cost may be higher in actual operation due to frequent replacement, high maintenance costs, rapid aging, and penalty losses. This improves the optimization of hybrid energy storage system capacity configuration.

[0031] The power imbalance correction factor in S5 includes: Through formula The power imbalance correction factor at time t is calculated. Where t is any time. Let be the actual power demand of the power grid at time t. This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t. Through the above technical solution, this example provides a method for obtaining the power imbalance correction factor, which can be achieved through the formula... The calculated correction factor reflects the deviation between the actual response power of the hybrid energy storage system and the grid demand. It is used for iterative coupling between the upper and lower level models. In a specific embodiment, during the two-level optimization process, the upper level model first decides the rated capacity and power allocation strategy of the energy storage system, and outputs the actual response power of the energy storage unit at time t. Then, the lower-level model determines the rated power and response strategy of the power-type energy storage based on this, and outputs the actual response power of the power-type energy storage unit at time t. And calculate the power imbalance correction factor at time t. ,when This indicates that the energy storage system's response is insufficient (actual output is less than demand), and the output of energy-type or power-type energy storage needs to be increased in the next iteration. This indicates an over-response (actual output exceeds demand), and the output needs to be reduced. If the response is normal, then there is no need to adjust the output of energy-type or power-type energy storage. By introducing the power imbalance correction factor in this way, the cooperative convergence problem in the two-layer decoupled optimization model is solved. This is achieved by adjusting the power imbalance correction factor at time t. Establish a feedback mechanism between the upper and lower layers: the lower layer will adjust the power imbalance correction factor at time t. Returning to the upper layer, the upper layer adjusts its decisions accordingly, including increasing energy storage capacity or adjusting power allocation strategies. After multiple iterations, the decisions of the upper and lower layers tend to be consistent, thereby improving iteration efficiency and the convergence success rate of the two-layer model.

[0032] The energy storage lifetime constraints in S6 include: The cycle life of energy storage units satisfies the following relationship: ; Where T is the total running time. Let be the depth of discharge at time t. This represents the maximum number of cycles that the energy storage unit can withstand under the deep discharge conditions corresponding to time t. The cycle life of power-type energy storage units satisfies the following relationship: The cumulative equivalent number of iterations shall not exceed the maximum allowed number of iterations. Through the above technical solution, this example provides energy storage lifetime constraints, including the cycle lifetime of the energy storage unit satisfying the following relationship: The cycle life of power-type energy storage units must satisfy the following relationship: the cumulative value of the equivalent cycle count does not exceed the maximum allowable cycle count. The charge-discharge half-cycles in its operating conditions can be counted using the rainflow counting method, and its cycle life constraint satisfies: ,in, This represents the total number of categories of different discharge depth cycles as determined by the rainflow counting method. This represents the total number of categories of different discharge depth cycles as determined by the rainflow counting method. For any discharge depth cycle category, For the first The actual number of deep discharge cycles that occurred. For the first Discharge depth corresponding to a type of discharge depth cycle For power-type energy storage units at a depth of discharge of The maximum allowed equivalent number of cycles is determined by fitting experimental data provided by energy storage unit manufacturers; a typical fitting model is... ,in, The maximum number of cycles to achieve 100% depth of discharge. The fitting coefficient is between 1.0 and 2.0. Its working principle is as follows: for energy-type energy storage units, a cumulative fatigue damage model is adopted. This constraint ensures that the cumulative fatigue damage caused by all charge and discharge cycles does not exceed the life limit of the energy storage unit. For power-type energy storage units, a simplified constraint is adopted that the cumulative value of the equivalent number of cycles does not exceed the maximum allowable number of cycles. With this setup, the lifetime constraint model based on cumulative fatigue damage accurately reflects the actual aging mechanism of the energy storage unit. Studies show that the damage caused by different discharge depths varies, therefore... Precisely quantifying the damage degree of a single cycle and accumulating it before comparing it with 1 is more accurate than the simple constraint of "the number of cycles does not exceed the rated value". Furthermore, the constraint and the lifetime loss cost complement each other. The lifetime constraint is a hard constraint (it must be met to ensure that the energy storage unit will not be prematurely scrapped within the planning period), while the loss cost is a soft constraint (it guides the optimization direction through cost, making the algorithm biased towards choosing low-damage strategies). The combination of the two ensures that the optimization result not only meets the basic lifetime requirements (hard constraints ensure feasibility) but also achieves optimal economic efficiency (soft constraints guide the optimization direction). This dual design of hard and soft constraints has a better optimization effect than simply using constraints or simply using cost.

[0033] The power balance constraints in S6 include: The following relationship must be satisfied:

[0034] in, This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t. The total power demand of the hybrid energy storage system at time t; Through the above technical solution, this example provides power balance constraints, including satisfying the following relationships: The constraint ensures that at any given time, the sum of the output power of the two energy storage units exactly equals the system demand, without any surplus or deficit. Specifically, the power balance constraint is a fundamental prerequisite for the normal operation of a hybrid energy storage system. If the power output does not match the demand, it can lead to poor frequency regulation (frequency deviation exceeds the dead zone) and failure of peak shaving and valley filling (peak values ​​are not smoothed out), or even grid frequency instability (frequency drop caused by power deficit) or voltage collapse (voltage rise caused by power surplus). Therefore, this basic engineering constraint is explicitly written into the optimization model to ensure the technical feasibility of the configuration scheme. Actual tests show that the configuration scheme that meets the power balance constraint achieves qualified frequency regulation. The efficiency improved from 78% to 96%, and the peak shaving and valley filling completion rate improved from 65% to 94%, demonstrating significant results. Furthermore, the power balance constraint tightly couples the power output of energy-type energy storage and power-type energy storage, reflecting the collaborative working characteristics of the hybrid energy storage system. The two types of energy storage units do not operate independently but jointly respond to the same demand signal. Through this constraint, the optimization model must simultaneously consider the capacity limitations of the two units (power upper limit, power lower limit, state of charge upper and lower limits, ramp rate limit, etc.) to find feasible power allocation schemes. This coupling constraint ensures that there is a feasible power allocation strategy under any operating condition, avoiding the imbalance problem caused by "each doing its own thing". Based on this, the constraint also forms a closed loop with the power imbalance correction factor, which is precisely the deviation of the power balance constraint. In the two-level iterative optimization, the correction factor drives the algorithm to continuously adjust the decision variables until the constraint is satisfied. (Approaching 0), it can be said that the power balance constraint is the goal and the correction factor is the means. Together, they ensure the feasibility of the final solution. This goal-means combination design makes the optimization process have both a clear goal orientation and an effective adjustment mechanism.

[0035] The state-of-charge constraints in S6 include: The state-of-charge constraints of energy storage units satisfy the following relationship: ; in, This is the lower limit threshold for the state of charge of the energy storage unit, with a value of 0.1. This is the upper limit threshold for the state of charge of the energy storage unit, with a value of 0.9. This represents the state of charge of the energy storage unit at time t. The state-of-charge constraints of power-type energy storage units satisfy the following relationship: ; in, This is the minimum threshold value for the state of charge of the power-type energy storage unit, taken as 0.05. This is the upper limit threshold for the state of charge of the power-type energy storage unit, with a value of 0.95. This represents the state of charge of a power-type energy storage unit at time t. Through the above technical solution, this example provides state of charge (SOC) constraints, including SOC constraints for energy-type energy storage units and power-type energy storage units. SOC constraints prevent overcharging or over-discharging of the energy storage unit, protecting its safety and reserving capacity for emergencies. SOC constraints are a necessary condition for ensuring the safe operation of the energy storage unit. Over-discharging (SOC below 0.1%) of energy-type energy storage such as lithium batteries can lead to electrode structure collapse, copper dissolution and deposition, causing irreversible capacity decay and internal short-circuit risks; overcharging (SOC above 0.1%)... 9) This can lead to lithium deposition and electrolyte decomposition, which may cause thermal runaway or even fire and explosion. In this example, the upper and lower limits of SOC for energy storage are set to 0.1 and 0.9 respectively, with a 20% safety buffer (10% upper limit + 10% lower limit), which effectively avoids safety risks. Although power storage (supercapacitor) has better safety, over-discharging (below 0.05) will cause the voltage to be too low and it will not work properly, and overcharging (above 0.95) will cause electrolyte decomposition. Therefore, by setting different thresholds, both types of energy storage units can operate within a safe range. Specifically, energy storage and power storage employ different SOC threshold ranges, reflecting the differences in technical characteristics between the two types of energy storage units. Energy storage (lithium batteries) is more sensitive to overcharging and over-discharging, requiring a wider safety margin. Meanwhile, lithium batteries have the best cycle life within the SOC range of 0.2-0.8; deviating from this range accelerates aging. Power storage (supercapacitors) has better overcharge and over-discharge tolerance and a wider usable SOC range (0-1). However, to reserve a certain amount of emergency backup capacity, boundaries of 0.05 and 0.95 are set. This invention's differentiated threshold settings ensure safety while maximizing the utilization of the energy storage unit's usable capacity. Compared to a scheme using a uniform threshold, it can further improve the usable capacity of power storage.

[0036] Please see Figure 2 As shown, the process of generating a configuration scheme in S7 includes: S71: Based on the output results of the rated capacity of the optimal energy storage unit, match the standard specification energy storage products available on the market to determine the battery type, series and parallel connection method, and number of battery clusters of the energy storage unit; S72: Based on the output results of the optimal rated capacity of the power energy storage unit, match the standard specifications of the power energy storage products available on the market to determine the energy storage type and the number of modules of the power energy storage unit; S73: Calculate system-level configuration parameters, including total system rated power, total system energy capacity, power ratio and energy ratio of energy-type energy storage and power-type energy storage, and configure electrical interface parameters, including DC bus voltage, number of PCS units and power allocation, step-up transformer capacity, and access bay location; S74: Based on the comprehensive life cycle cost model, it outputs economic indicators such as initial investment, annual operation and maintenance cost, and levelized cost of energy storage. S75: Based on the monitoring point locations and power grid topology in S1, output installation location suggestions; S76: Generate a structured configuration scheme that includes the overall system configuration, details of energy storage units, electrical interface parameters, control strategies, economic indicators, and installation locations; Through the above technical solution, this example provides the process of generating a configuration scheme in S7. By combining the output results of the rated capacity of the optimal energy-type energy storage unit and the output results of the rated capacity of the optimal power-type energy storage unit, standard-specification power-type energy storage products and standard-specification energy-type energy storage products available on the market can be matched. Then, based on the system-level configuration parameters, the configuration of electrical interface parameters can be realized. Finally, by summarizing the comprehensive cost model of the entire life cycle and the installation configuration suggestions, the optimized configuration of the grid-side hybrid energy storage system capacity can be generated, thereby improving the operational stability of the power grid. It should be noted that the recommended installation locations for outputs in S75 include: calculating the short-circuit capacity and voltage support sensitivity of each candidate access point in the power grid; recommending that energy-type energy storage units be configured near monitoring points in areas with low short-circuit capacity or heavy loads to provide effective peak shaving and valley filling and voltage support; recommending that power-type energy storage units be configured near renewable energy collection stations or areas sensitive to frequency fluctuations to ensure rapid response to high-frequency fluctuations; and when the optimal capacity configuration is large and multiple access points are required, the specific capacity should be determined according to the principle of allocating according to the demand intensity ratio of each monitoring point. It should be noted that voltage support sensitivity can be calculated using the formula... Calculated, where, The reactive power change at the injected candidate access point is expressed in megavars (MVar). The voltage change at this node is the amount of change caused by the change in reactive power, and the unit is kilovolt (kV); a low short-circuit capacity is defined as a short-circuit capacity of less than or equal to 500 megavolt-amperes; It should be noted that the demand intensity within the specific capacity is determined by the principle of allocating it according to the proportion of demand intensity at each monitoring point, using the formula... The calculation yields the result, where m is any monitoring point. Let m be the total number of scene events associated with the m-th monitoring point. For any scene event, The m-th monitoring point is associated with the first... Power deviation amplitude in each scene event For all The maximum value in.

[0037] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for optimizing the capacity configuration of grid-side hybrid energy storage systems for multiple application scenarios, characterized in that, The method includes: S1: Collect historical operating data from at least one monitoring point on the power grid side within a sampling period according to a fixed sampling time interval, and construct a time series sample set after normalization and outlier removal. The historical operating data includes active power, frequency deviation and voltage deviation. S2: Extract multi-dimensional feature vectors based on time series sample sets, and use clustering algorithms to divide the power grid operation status into multiple application scenarios, including frequency regulation, peak shaving and valley filling, fluctuation smoothing, and reserve support. Output the scenario label corresponding to each sampling time. ; S3: Establish a hybrid energy storage system topology model, including power-type energy storage units and energy-type energy storage units, and establish a comprehensive life cycle cost model, which includes purchase cost, replacement cost, operation and maintenance cost, loss cost and penalty cost; S4: Construct adaptive weight optimization models for different scenarios, and dynamically adjust the economic target weight, reliability target weight, and response speed target weight based on different scenario labels; S5: Establish a two-layer decoupled optimization model for power and energy. The upper-layer model uses the rated capacity of the energy-type energy storage unit as the decision variable to perform peak shaving and valley filling or reserve capacity matching. The lower-layer model uses the rated power of the power-type energy storage unit as the decision variable to perform high-frequency fluctuation suppression and frequency regulation command tracking. The upper and lower layers are iteratively coupled through the power imbalance correction factor. S6: The two-layer model described in S5 is solved using an optimization algorithm. Under the conditions of satisfying power balance constraints, state of charge constraints, and energy storage lifetime constraints, the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit are output. S7: By combining the optimal rated capacity of the energy-type energy storage unit and the rated power of the power-type energy storage unit output in S6, a hybrid energy storage system capacity configuration scheme is generated, which includes energy storage type, rated capacity, rated power, and installation location recommendations.

2. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios as described in claim 1, characterized in that, The multidimensional feature vector extracted in S2 includes: The volatility index, maximum power change rate, and energy gap duration feature vectors extracted from active power data in a time series sample set; Feature vectors of frequency over-limit amplitude and frequency over-limit duration extracted from frequency deviation data; Feature vectors of voltage over-limit amplitude and voltage over-limit duration extracted from voltage deviation data.

3. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios as described in claim 1, characterized in that, The process of dynamically adjusting the target weight in S4 includes: By combining the normalized demand intensity of economic, reliability, and response speed objectives under different scenario labels, the weights of these objectives are dynamically adjusted.

4. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios as described in claim 1, characterized in that, The process of establishing the power and energy two-layer decoupled optimization model in S5 includes: A first-order low-pass filtering method is used to allocate power between power-type energy storage units and energy-type energy storage units. The filtering time constant is adaptively determined by the scenario label. A second-level time constant is used in frequency modulation and fluctuation smoothing scenarios, while a minute-level time constant is used in peak shaving and valley filling scenarios and backup support scenarios, in order to distinguish the allocation paths of high-frequency power and low-frequency power.

5. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios according to claim 1, characterized in that, The process of establishing the comprehensive lifecycle cost model in S3 includes: The total lifecycle cost is calculated by combining the initial purchase cost, replacement cost, operation and maintenance cost, life loss cost, and penalty cost.

6. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios as described in claim 1, characterized in that, The power imbalance correction factor in S5 includes: Through formula The power imbalance correction factor at time t is calculated; where t is any time. Let be the actual power demand of the power grid at time t. This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t.

7. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios as described in claim 1, characterized in that, The energy storage lifetime constraints in S6 include: The cycle life of energy storage units satisfies the following relationship: ； Where T is the total running time. Let be the depth of discharge at time t. This represents the maximum number of cycles that the energy storage unit can withstand under the deep discharge conditions corresponding to time t. The cycle life of power-type energy storage units satisfies the following relationship: The cumulative value of the equivalent loop count shall not exceed the maximum allowed loop count.

8. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios according to claim 1, characterized in that, The power balance constraints in S6 include: The following relationship must be satisfied: ； in, This represents the actual response power of the energy storage unit at time t. This represents the actual response power of the power-type energy storage unit at time t. Let be the total power demand of the hybrid energy storage system at time t.

9. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios according to claim 1, characterized in that, The state-of-charge constraints in S6 include: The state-of-charge constraints of energy storage units satisfy the following relationship: ； in, This is the lower limit threshold for the state of charge of the energy storage unit, with a value of 0.

1. This is the upper limit threshold for the state of charge of the energy storage unit, with a value of 0.

9. This represents the state of charge of the energy storage unit at time t. The state-of-charge constraints of power-type energy storage units satisfy the following relationship: ； in, This is the minimum threshold value for the state of charge of the power-type energy storage unit, taken as 0.

05. This is the upper limit threshold for the state of charge of the power-type energy storage unit, with a value of 0.

95. This represents the state of charge of a power-type energy storage unit at time t.

10. The method for optimizing the capacity configuration of a grid-side hybrid energy storage system for multiple application scenarios according to claim 1, characterized in that, The process of generating the configuration scheme in S7 includes: S71: Based on the output results of the rated capacity of the optimal energy storage unit, match the standard specification energy storage products available on the market to determine the battery type, series and parallel connection method, and number of battery clusters of the energy storage unit; S72: Based on the output results of the optimal rated capacity of the power energy storage unit, match the standard specifications of the power energy storage products available on the market to determine the energy storage type and the number of modules of the power energy storage unit; S73: Calculate system-level configuration parameters, including total system rated power, total system energy capacity, power ratio and energy ratio of energy-type energy storage and power-type energy storage, and configure electrical interface parameters, including DC bus voltage, number of PCS units and power allocation, step-up transformer capacity, and access bay location; S74: Based on the comprehensive life cycle cost model, it outputs economic indicators such as initial investment, annual operation and maintenance cost, and levelized cost of energy storage. S75: Based on the monitoring point locations and power grid topology in S1, output installation location suggestions; S76: Generate a structured configuration scheme that includes the overall system configuration, details of energy storage units, electrical interface parameters, control strategies, economic indicators, and installation locations.

Images