Hybrid energy storage method based on distributed photovoltaic power generation

Abstract

The invention relates to the technical field of photovoltaic power generation, in particular to a hybrid energy storage method based on distributed photovoltaic power generation. According to the method, a degradation cost function of a storage battery and a super capacitor is established, photovoltaic output and electric vehicle charging demand prediction are combined, a multi-target model prediction control algorithm is constructed, and an optimal charging and discharging instruction for minimizing the total operation cost and maximizing the photovoltaic power generation utilization rate is calculated. The instruction is decomposed into a high-frequency component and a low-frequency component through wavelet transformation, the high-frequency component and the low-frequency component are distributed to a super capacitor and a storage battery respectively, and power output is adjusted in real time through a local controller. According to the method, the dynamic fitness function is constructed according to the long-term cost value and the operation data, and the photovoltaic and energy storage unit capacity configuration scheme with the maximum economic benefit is solved by using the particle swarm optimization, so that high-dynamic load scenes such as a charging station and the like are effectively coped with while the long-term economic benefit and the stability are ensured.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power generation technology, specifically to a hybrid energy storage method based on distributed photovoltaic power generation. Background Technology

[0002] Due to the intermittent and unstable nature of solar energy, distributed photovoltaic (PV) power generation systems often experience peak-shaving phenomena, leading to curtailment and unstable power supply. Combining hybrid energy storage technology with distributed PV power generation significantly solves this problem. In electric vehicle charging station scenarios, the frequent power surges caused by voltage fluctuations are particularly prominent. Against this backdrop, developing an efficient, stable, and economical hybrid energy storage method faces a series of technical challenges.

[0003] Existing hybrid energy storage methods have several drawbacks: Some power allocation methods simply allocate high-frequency fluctuations to supercapacitors and low-frequency and DC components to batteries based on preset cutoff frequencies. When the cutoff frequency of the filter is not set properly, power will frequently switch between high and low frequencies, causing drastic power fluctuations that affect control accuracy and even the safe and stable operation of the power grid. Some methods do not fully consider the state of charge of the battery and the voltage level of the supercapacitor. When the battery's state of charge is too high, charging power is still forcibly allocated to it, triggering its internal protection mechanism, causing the charging power to suddenly drop to zero and all excess power to be instantly transferred to the supercapacitor, resulting in drastic power fluctuations. At the same time, inaccurate control forces the battery to undergo violent charging and discharging, which exacerbates its performance degradation, severely shortens battery life, and increases operating costs. Inappropriate energy storage capacity configuration prevents the system from fully absorbing excess power, leading to curtailment of solar power and reduced energy utilization efficiency. The lag in capacity configuration, in turn, exacerbates the frequent high-intensity cycling of the battery within its limited capacity range, causing secondary damage to battery health and unnecessary economic losses. In summary, existing systems struggle to balance stable operation and battery life protection, resulting in high operating costs and poor long-term economic benefits.

[0004] To address this, a hybrid energy storage method based on distributed photovoltaic power generation is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a hybrid energy storage method based on distributed photovoltaic (PV) power generation. This invention establishes degradation cost functions for batteries and supercapacitors, and combines PV output and electric vehicle charging demand forecasts to construct a multi-objective model predictive control algorithm. This algorithm calculates the optimal charging and discharging command that minimizes total operating costs and maximizes PV power generation utilization. This command is decomposed into high-frequency and low-frequency components using wavelet transform and allocated to the supercapacitor and battery respectively, with the power output adjusted in real time by a local controller. A dynamic fitness function is constructed based on long-term cost values ​​and operating data. A particle swarm optimization algorithm is used to solve for the PV and energy storage unit capacity configuration scheme that maximizes economic benefits, thereby effectively addressing high-dynamic load scenarios such as charging stations while ensuring the long-term economic benefits and stability of the system.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A hybrid energy storage method based on distributed photovoltaic power generation includes:

[0008] A battery degradation cost function is established based on the charge / discharge rate, power surge frequency, and pulse duration. A hybrid energy storage degradation cost function is established by combining the supercapacitor degradation cost function, and the long-term cost value is calculated.

[0009] A multi-objective model predictive control algorithm is constructed to predict the uncertain range of photovoltaic output based on historical data and meteorological data, and to predict the load distribution range based on user behavior pattern data. Based on the long-term cost value, the uncertain range of photovoltaic output, and the load distribution range, the charging and discharging power command that minimizes the expected value of total operating cost and maximizes the utilization rate of photovoltaic power generation is calculated and decomposed into high-frequency and low-frequency components through wavelet transform.

[0010] The real-time state of charge of the energy storage unit is obtained. Based on the real-time state of charge and the preset safety threshold, the local controller autonomously adjusts the charging and discharging power output, allocating high-frequency components to the supercapacitor and low-frequency components to the battery.

[0011] Obtain operational data and long-term cost values, construct a dynamic fitness function, and use the particle swarm optimization algorithm to solve for the optimal economic benefit of photovoltaic and energy storage unit capacity configuration schemes; then configure the capacity of photovoltaic and energy storage units for the next cycle.

[0012] Preferably, the implementation method of the battery degradation cost function includes: establishing a mapping relationship between the actual cycle life of the battery and the battery health status under different discharge depths and charge / discharge rates based on historical operating data; quantifying the additional losses to the battery caused by the power surge brought by electric vehicle charging by using charge / discharge rate, power surge frequency and pulse duration as variables, and calculating the economic losses caused to battery life by different charge / discharge behaviors by combining the battery's discharge depth, health status and unit capacity investment cost;

[0013] The specific implementation of the supercapacitor degradation cost function includes: collecting historical operating data and counting the actual number of cycles of the supercapacitor in charging and discharging scenarios; establishing a linear degradation model by correlating the number of cycles with the unit power investment cost of the supercapacitor, and calculating the economic losses of the supercapacitor in frequently coping with high-frequency fluctuations in charging load.

[0014] Preferably, historical photovoltaic power output data, meteorological data, and charging power demand data of charging piles are collected, as well as user behavior pattern data of charging stations. After preprocessing all collected data, feature engineering is performed to capture the periodicity and randomness of charging load. The uncertainty range of photovoltaic power output is predicted through long short-term memory network, and the load distribution range of electricity load is predicted through gradient boosting tree. User behavior patterns include vehicle access time, charging duration, and charging amount.

[0015] Preferably, a multi-objective optimization function is constructed using the predicted uncertain range of photovoltaic output, the load distribution range, and the degradation cost function as inputs. The safety thresholds of energy storage unit power balance and state of charge are used as constraints. In each control cycle, stochastic optimization is performed based on the predicted uncertainty range to generate the expected total operating cost that minimizes the total operating cost and maximizes the photovoltaic power generation utilization rate. The minimized total operating cost includes electricity purchase and sale cost, hybrid energy storage degradation cost, and operation and maintenance cost. The constraints include power balance constraints, energy storage unit constraints, and power upper and lower limit constraints.

[0016] Preferably, the supercapacitor local controller uses current inner loop control to precisely adjust the charging and discharging current, quickly respond to and absorb high-frequency power components, and smooth out instantaneous power surges; the battery local controller uses voltage and current dual closed-loop control to smoothly handle low-frequency power components.

[0017] Wavelet transform is used to process the charging and discharging power command output by the multi-objective model predictive control algorithm, decomposing the command into low-frequency and high-frequency components. The low-frequency component represents the long-term energy throughput demand of the system and is used to guide the stable charging and discharging of the battery. The high-frequency component represents the instantaneous power fluctuation of the system and is used to guide the rapid response of the supercapacitor.

[0018] The local controller autonomously adjusts its charging and discharging rate and power output, including: when the state of charge of the battery exceeds the preset safe operating range, the battery local controller automatically limits the charging and discharging power; the supercapacitor local controller limits the charging and discharging current when the real-time voltage and temperature exceed the safety threshold.

[0019] Preferably, the dynamic fitness function balances the costs of purchasing and selling electricity, maintenance costs, and hybrid energy storage degradation costs with peak shaving and valley filling benefits and grid ancillary service revenues, and calculates the maximum economic benefits using a particle swarm optimization algorithm; it performs stochastic optimization based on predicted load intervals to output a capacity configuration scheme that extends battery life and meets the stochastic load demands of charging stations; the operational data includes historical aggregated state of charge data, energy utilization rate, battery health status, and grid interaction benefits.

[0020] Preferably, the capacity configuration for the photovoltaic and energy storage units in the next cycle includes reserving redundant capacity for the future growth of the number of electric vehicles in the capacity configuration scheme; updating the rated capacity of the photovoltaic power generation unit and the rated capacity of the energy storage unit according to the optimal capacity configuration scheme; and sending the updated capacity parameters to the control logic of each hierarchical controller as new operating constraints.

[0021] Preferably, the central manager converts the capacity configuration scheme into macroscopic operating parameters and strategy instructions; the regional controller receives the macroscopic operating parameters and strategy instructions, calculates the optimal charging and discharging power instructions through a multi-objective model predictive control algorithm, and decomposes the power instructions into high-frequency and low-frequency components, and issues the component instructions as execution instructions; each energy storage unit and charging pile has a local controller that receives the component instructions, combines them with real-time operating data, and autonomously adjusts its charging and discharging rate and power output, allocating high-frequency component instructions to the supercapacitor and low-frequency component instructions to the battery, and performing safety correction when real-time data exceeds a preset safety threshold.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. By establishing a hybrid energy storage degradation cost function, key parameters such as battery charge / discharge rate and health status are quantified into economic costs. Combined with multi-objective model predictive control algorithm for optimization, the contradiction between high-frequency charging and discharging and battery life in charging station scenarios is resolved. This not only achieves short-term economic benefits such as peak shaving and valley filling, but also ensures that the total operating cost of the system is minimized and the long-term economic benefits are maximized by incorporating battery degradation costs into the calculation.

[0024] 2. A hierarchical collaborative control architecture is adopted to decompose complex control tasks into different levels. Long short-term memory networks and gradient boosting trees are used to predict photovoltaic output and charging load to provide forward-looking planning for the system. When there is a deviation between the actual load and the prediction, the lower-level local controller can allocate instantaneous impact and stable tasks to the supercapacitor and the battery respectively according to the real-time state of charge and the high-frequency and low-frequency commands decomposed by wavelet transform. This hierarchical collaboration and precise response mechanism ensures the stable operation of the system under high dynamic and high random loads and effectively smooths out power fluctuations.

[0025] 3. Based on aggregated state of charge data, energy utilization rate, and grid interaction benefits, a dynamic fitness function is constructed, and the optimal capacity configuration scheme is solved using the particle swarm optimization algorithm. This enables the system's capacity planning to be dynamically optimized based on historical data and long-term operational goals, and reserves redundant capacity for future growth in the number of electric vehicles. It also provides optimization suggestions for system managers, which not only improves the rationality of system configuration but also provides intelligent decision support for the efficient operation of charging stations. Attached Figure Description

[0026] Figure 1 A flowchart of a hybrid energy storage method based on distributed photovoltaic power generation provided in an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of a distributed control method based on a hybrid energy storage system for distributed photovoltaic power generation, provided as an embodiment of the present invention.

[0028] Figure 3 This is a schematic diagram of the multi-objective model prediction algorithm structure provided in an embodiment of the present invention. Detailed Implementation

[0029] 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.

[0030] Please see Figures 1 to 3 This invention provides a hybrid energy storage method based on distributed photovoltaic power generation, the technical solution of which is as follows:

[0031] A hybrid energy storage method based on distributed photovoltaic power generation includes establishing a battery degradation cost function based on charge / discharge rate, power impulse frequency and pulse duration, establishing a hybrid energy storage degradation cost function in combination with the supercapacitor degradation cost function, and calculating the long-term cost value;

[0032] A multi-objective model predictive control algorithm is constructed to predict the uncertain range of photovoltaic output based on historical data and meteorological data, and to predict the load distribution range based on user behavior pattern data. Based on the long-term cost value, the uncertain range of photovoltaic output, and the load distribution range, the charging and discharging power command that minimizes the expected value of total operating cost and maximizes the utilization rate of photovoltaic power generation is calculated and decomposed into high-frequency and low-frequency components through wavelet transform.

[0033] The real-time state of charge of the energy storage unit is obtained. Based on the real-time state of charge and the preset safety threshold, the local controller autonomously adjusts the charging and discharging power output, allocating high-frequency components to the supercapacitor and low-frequency components to the battery.

[0034] Obtain operational data and long-term cost values, construct a dynamic fitness function, and use the particle swarm optimization algorithm to solve for the optimal economic benefit of photovoltaic and energy storage unit capacity configuration schemes; then configure the capacity of photovoltaic and energy storage units for the next cycle.

[0035] Example 1:

[0036] This embodiment applies to an electric vehicle charging station that integrates distributed photovoltaic power generation. The station has 50 charging piles and is equipped with a 1.5MWp distributed photovoltaic power generation system and a hybrid energy storage system. The hybrid energy storage system consists of a 1MWh / 500kW lithium iron phosphate battery pack and a 100kW / 20kWh supercapacitor pack.

[0037] As one embodiment of the present invention, refer to Figure 1 A flowchart of a hybrid energy storage method based on distributed photovoltaic power generation, referring to... Figure 2 This is a schematic diagram of a distributed control method based on distributed photovoltaic power generation and hybrid energy storage, referring to... Figure 3 This is a schematic diagram of the multi-objective model prediction algorithm structure.

[0038] Furthermore, the implementation of the battery degradation cost function includes: establishing a mapping relationship between the actual cycle life of the battery and the battery health status under different discharge depths and charge / discharge rates based on historical operating data; quantifying the additional losses to the battery caused by the high power surge brought by electric vehicle charging by using charge / discharge rate, power surge frequency, and pulse duration as key variables; and calculating the economic losses to battery life caused by different charge / discharge behaviors by combining the battery's discharge depth, health status, and unit capacity investment cost.

[0039] The specific implementation of the supercapacitor degradation cost function includes: collecting historical operating data and counting the actual number of cycles of the supercapacitor in high-frequency, high-power charging and discharging scenarios; establishing a linear degradation model by correlating the number of cycles with the unit power investment cost of the supercapacitor, and calculating the economic losses of the supercapacitor in frequently coping with high-frequency fluctuations in charging load.

[0040] Specifically, long-term monitoring of an energy storage unit composed of lithium iron phosphate batteries (500 kWh capacity, unit capacity investment cost of 1000 yuan / kWh) was conducted, collecting operational data under high-power fast charging scenarios. Analysis showed that when the charging rate increased from 0.5C to 2C, the battery cycle life dropped sharply from approximately 3000 cycles to 1500 cycles, and the rate of degradation in healthy state accelerated by 50%. By combining this data with the unit capacity investment cost, the economic losses to battery life caused by different charge / discharge behaviors can be calculated. For example, at a 2C rate, the life loss cost per complete cycle is 333.3 yuan / cycle, while at a 0.5C rate, the life loss cost is 166.7 yuan / cycle. The additional losses caused by high-rate charge / discharge are quantified using a hybrid energy storage degradation cost function, providing a reliable optimization basis for upper-level control algorithms. The supercapacitor has a rated power of 200kW and a unit power investment cost of 2000 yuan / kW. It can cope with more than 1000 instantaneous power fluctuations of power pulses per day on average. By establishing a linear degradation model, the number of cycles is correlated with the unit power investment cost to calculate the economic loss when coping with high-frequency fluctuations.

[0041] This method quantifies the additional damage to the battery caused by high-rate charging and discharging, and more accurately assesses the economic losses caused by different operating strategies to battery life. It provides a reliable optimization basis for the upper-level control algorithm, and effectively extends the service life of the battery and supercapacitor while avoiding excessive damage.

[0042] Furthermore, historical photovoltaic power output data, meteorological data, charging power demand data of charging piles, and user behavior pattern data of charging stations are collected. All collected data are preprocessed and feature engineering is used to capture the periodicity and randomness of charging load. Long short-term memory networks and gradient boosting trees are used to predict the uncertainty range of photovoltaic power output and the load distribution range of electricity consumption. A charging load prediction algorithm is constructed based on the charging behavior patterns of historical data. By clustering user groups according to charging behavior characteristics and establishing independent load prediction models for different user groups, the prediction of charging load is dynamically adjusted according to the real-time access of electric vehicle behavior characteristics.

[0043] This method can more accurately predict the random load characteristics of charging stations by classifying and modeling user behavior patterns, and adjust the energy storage scheduling strategy in advance to avoid resource waste or response delays due to inaccurate predictions, thereby improving the overall operating efficiency and robustness of the system.

[0044] Specifically, historical photovoltaic power output data of electric vehicle charging stations is acquired, along with meteorological data including solar radiation intensity, ambient temperature, cloud cover, and wind speed. Historical charging power demand data for charging piles within the charging stations is also acquired, along with historical user charging behavior data including vehicle access time, charging duration, and charging amount. The collected data is cleaned, normalized, and outlier-handled. Subsequently, feature engineering is performed to capture the periodicity and randomness of the charging load. Periodicity features are various time-point characteristics in the timestamps, such as significantly higher charging demand during weekday morning and evening peak hours, significantly higher charging demand from Monday to Friday than on Saturday and Sunday, and significantly higher charging demand in winter than in summer. Randomness features are extracted from random user behavior, such as significant differences in charging behavior on specific holidays compared to normal times.

[0045] Long Short-Term Memory (LSTM) networks predict the uncertainty range of photovoltaic (PV) output within the next 24 hours based on historical PV output data and meteorological data. For example, the model predicts that the PV output from 13:00 to 14:00 on a given day will be 850±50kW, indicating that the actual PV output will fall between 800kW and 900kW within a 95% confidence interval. Historical charging load data, user behavior pattern data, and periodic characteristics are input into the gradient boosting tree model to predict the electricity load distribution range for each time period within the next 24 hours. For example, based on historical user behavior data, the model predicts that the load distribution range during the evening peak from 19:00 to 20:00 will be 1200±150kW. This prediction range reflects the randomness of user behavior, enhances the robustness of the system, and ensures that the energy storage system can plan charging and discharging strategies in advance to cope with potential load shocks.

[0046] By employing long short-term memory networks and gradient boosting trees to predict the uncertainty range of photovoltaic power output and charging load, the system provides forward-looking decision support, enabling the system to plan charging and discharging strategies in advance to effectively cope with the randomness of photovoltaic power output and the high dynamism of charging load.

[0047] Furthermore, a multi-objective optimization function is constructed using the predicted photovoltaic output, load distribution range, and degradation cost function as inputs. The power balance and state of charge safety thresholds of the energy storage units are used as constraints. In each control cycle, stochastic optimization is performed based on the predicted uncertainty range to generate the expected total operating cost that minimizes the total operating cost and maximizes the photovoltaic power generation utilization rate. The minimized total operating cost includes the cost of purchasing and selling electricity, the degradation cost of hybrid energy storage, and the operation and maintenance cost. The constraints include power balance constraints, energy storage unit constraints, and upper and lower power limits constraints.

[0048] Using predicted photovoltaic (PV) output, load distribution range, and degradation cost function as inputs, a multi-objective optimization function is constructed to minimize total operating cost and maximize PV power generation utilization. Minimizing total operating cost is the minimum expected value of the sum of electricity purchase and sale costs, degradation costs, and operation and maintenance costs. Maximizing PV power generation utilization is the maximum value of the relationship between actual PV power generation and total PV power generation. When solving the optimization function, power balance and energy storage unit constraints must be satisfied to ensure safe and stable system operation. The power balance constraint is the balance between PV output, grid interaction power, charging pile load, and the charging and discharging power of the hybrid energy storage system. The energy storage unit constraint is that the battery's state of charge (SOC) must be maintained within the safe operating range. min ≤SOC bat (t)≤SOC max SOC min This indicates that the minimum state of charge (SOC) is set to 15%. max The maximum state of charge (SOC) is set to 95%. If the predicted SOC exceeds this range, the optimization algorithm will automatically adjust the charging and discharging power to ensure that the charging and discharging power of the battery and supercapacitor are within their rated power range. By combining real-time acquisition of price signals from the grid ancillary services market with minimizing total operating costs, a dynamic optimization model oriented towards the electricity market is constructed to solve for maximizing economic benefits and dynamically adjusting the charging and discharging strategy of the hybrid energy storage system. While meeting local load demand, it participates in grid ancillary services as a virtual power plant. This method transforms charging stations from passive electricity loads into active electricity market participants by responding to grid demand in real time and utilizing their flexibility to obtain additional revenue in the ancillary services market. This not only brings more substantial long-term economic returns to charging stations but also enhances the stability and reliability of the grid.

[0049] Specifically, the model predicts that the photovoltaic output will be 850±50kW and the charging load demand will be 700±100kW during a certain period. At this time, the battery discharges at a rate higher than 1.5C, using a 15-minute control cycle. Based on the predicted photovoltaic output and charging load demand, a stochastic optimization model is established to balance long-term economic benefits and photovoltaic power generation utilization to calculate the optimal charging and discharging command. The command instructs the battery to charge smoothly at a rate of 0.5C and store the excess 150kW of photovoltaic power. Under the premise of ensuring power balance, the system obtains about 120 yuan in electricity revenue through peak shaving and valley filling. When adopting the smooth charging strategy, the degradation cost of the hybrid energy storage system is only about 50 yuan, which is far lower than the potential losses under the high-rate charging scenario. For example, the degradation cost of charging at a rate of 2C is as high as 100 yuan or more. This method constructs a multi-objective stochastic optimization function by predicting photovoltaic output and charging load demand to balance long-term economic benefits and photovoltaic power generation utilization, ensuring that the system maintains efficient and stable operation while pursuing revenue.

[0050] Furthermore, the supercapacitor local controller uses current inner loop control to precisely adjust the charging and discharging current, quickly respond to and absorb high-frequency power components to smooth out instantaneous power surges; the battery local controller uses voltage and current dual closed-loop control to smoothly handle low-frequency power components to avoid losses caused by high-frequency fluctuations.

[0051] Wavelet transform is used to decompose the charging and discharging power command output by the multi-objective model predictive control algorithm into low-frequency and high-frequency components. The wavelet transform is Daubechies wavelet, and the number of decomposition layers is set to 4 based on the characteristics of historical load data. This effectively separates the low-frequency component representing long-term energy demand from the instantaneous high-frequency component representing instantaneous power fluctuations to ensure the accuracy of power allocation. When the state of charge of the battery exceeds the preset safe operating range, the battery local controller automatically limits its charging and discharging power. The supercapacitor local controller limits its charging and discharging current based on real-time voltage, temperature, and the set safety threshold.

[0052] A charging pile-local controller collaborative power allocation strategy is adopted. The local controller of each charging pile communicates in real time with the local controller of the hybrid energy storage unit based on the battery health status and charging demand of the connected vehicle and coordinates to smooth the charging power.

[0053] This method decentralizes power allocation decisions from the central level to the execution terminal to achieve more refined coordination. Through real-time communication between charging piles and energy storage units, power can be intelligently allocated according to actual needs, avoiding power spikes caused by multiple fast charging demands occurring simultaneously, thereby effectively mitigating the impact on the power grid.

[0054] The charging pile and the local controller of the energy storage unit communicate in real time via the Modbus TCP protocol. When the total power demand exceeds 90% of the rated power of the energy storage system, the local controller of the charging pile dynamically adjusts the charging current according to the real-time charging demand of each vehicle and the real-time state of charge and maximum discharge power of the energy storage unit to ensure that the total power demand does not exceed the system limit.

[0055] Specifically, when 20 electric vehicles simultaneously access the charging station for fast charging, the total load suddenly surges from 400kW to 1200kW, accompanied by frequent power pulses over the next 15 minutes. The low-frequency component obtained by wavelet transform decomposition of this power command sequence remains at approximately 700kW during the period from 14:00 to 14:15. This portion of the power is directed to be allocated to the battery for stable power supply to the load. The high-frequency component obtained after decomposition appears as multiple positive and negative pulses with peak values ​​of ±300kW during the aforementioned period. This portion of the power is directed to be allocated to the supercapacitor.

[0056] When a momentary power pulse occurs due to fast charging of an electric vehicle, the supercapacitor local controller quickly adjusts the charging and discharging current within 50 milliseconds to absorb the high-frequency power, thereby protecting the battery from the impact. The controller monitors the voltage and temperature of the supercapacitor in real time. When the voltage exceeds the safety threshold of 2.7V / cell or the temperature exceeds 60℃, it immediately limits the charging and discharging current to prevent overvoltage or overheating and ensure equipment safety.

[0057] By using wavelet transform to decompose the power command into high-frequency and low-frequency components and distributing them to the supercapacitor and battery respectively, precise power coordination is achieved. This effectively mitigates the instantaneous power surge caused by high dynamic loads, protects the battery from unnecessary losses, and significantly improves the system's response speed and stability.

[0058] Furthermore, the dynamic fitness function weighs the costs of purchasing and selling electricity, maintenance costs, hybrid energy storage degradation costs, peak shaving and valley filling benefits, and grid ancillary service revenues, and calculates the maximum economic benefits using the particle swarm optimization algorithm; based on the predicted load range, risk-based optimization is performed to enhance the robustness of the configuration scheme in response to potential load shocks, and outputs a capacity configuration scheme that extends battery life and meets the random load demand of charging stations.

[0059] The dynamic fitness function is a comprehensive indicator designed to weigh the long-term economic benefits of the system. It considers not only the costs of electricity purchase and sale, maintenance, and the degradation costs of the energy storage system, but also the benefits gained through peak shaving and valley filling and participation in grid ancillary services. This function quantifies these complex economic indicators, providing a clear optimization objective for the particle swarm optimization (PSO) algorithm. In the PSO algorithm's solution process, each particle represents a configuration scheme for photovoltaic and energy storage capacity and iterates continuously. The fitness function evaluates the merits of each scheme, ultimately converging to an optimal configuration that maximizes the overall economic benefits of the system.

[0060] Specifically, historical data shows that the current system often experiences solar power curtailment during midday in summer, with energy utilization rate at only 80%. During evening peak hours, the interaction power between charging load and the grid frequently exceeds 1000kW, resulting in high electricity costs. The unit degradation cost of batteries under high-power cycling is 0.33 yuan / kWh. By using the particle swarm optimization algorithm, when evaluating each particle, the algorithm considers the upper and lower limits of the predicted load range, rather than a single predicted value. For example, for a configuration scheme of [1MW, 2MWh], if the predicted load during a certain evening peak may reach 1200kW, the algorithm will evaluate the robustness of the scheme in dealing with this high load impact and the possible additional electricity purchase costs or degradation costs.

[0061] After multiple iterations, the algorithm converged and output the optimal capacity configuration scheme: the rated photovoltaic capacity is 1.5MW and the rated energy storage capacity is 2MWh. This scheme is expected to reduce the annualized operating cost by 30%, increase the energy utilization rate to over 95%, meet the random load demand of the charging station, and extend the battery life to maximize the overall economic benefits.

[0062] This method uses the particle swarm optimization algorithm to dynamically optimize the capacity, so that the capacity configuration of photovoltaic and energy storage is no longer static but can be dynamically adjusted according to historical operating data and long-term economic goals. This improves the rationality of system configuration and enhances the robustness to cope with future load changes.

[0063] Furthermore, the capacity configuration for the photovoltaic and energy storage units in the next cycle is carried out, including reserving redundant capacity for the future growth of electric vehicle charging stations in the capacity configuration scheme; updating the rated capacity of the photovoltaic power generation units and the rated capacity of the energy storage units according to the optimal capacity configuration scheme; and sending the updated capacity parameters to the control logic of each hierarchical controller as new operating constraints to guide the power scheduling and charging and discharging management in the next cycle.

[0064] Specifically, the peak load of charging stations is predicted to increase by 20% to 1200kW. To smoothly cope with this growth, the capacity configuration scheme will reserve 20% of redundant capacity. The algorithm will suggest expanding the energy storage capacity to 2.4MWh and the photovoltaic capacity to 1.8MW in stages over two years.

[0065] After receiving new macroscopic operating parameters, the area controller will adjust the upper limit of the rated capacity of the energy storage unit from 1.5MWh to 2MWh and the upper limit of the photovoltaic capacity from 1MW to 1.5MW when executing the multi-objective model predictive control algorithm. After receiving the new rated capacity parameters, the local controller of each energy storage unit will adjust its internal power upper and lower limit constraints and state-of-charge safety threshold. For example, if the energy storage unit capacity is updated to 2MWh, the energy range corresponding to its state-of-charge safety operating range (20% to 90%) will also be expanded accordingly to guide the local controller to perform charging and discharging management more flexibly.

[0066] The system's forward-looking approach, which reserves redundant capacity for future growth in the number of electric vehicles, ensures that charging stations can smoothly cope with market growth and avoid operational bottlenecks and equipment damage caused by insufficient capacity, thus providing a guarantee for sustainable operation.

[0067] Furthermore, the central management system transforms the capacity configuration scheme into macroscopic operating parameters and strategy instructions; the regional controller receives the macroscopic operating parameters and strategy instructions, calculates the optimal charging and discharging power instructions through a multi-objective model predictive control algorithm, decomposes the power instructions into high-frequency and low-frequency components, and issues the component instructions as execution instructions; each energy storage unit and charging pile has a local controller that receives the component instructions, combines real-time operating data to autonomously adjust its charging and discharging rate and power output, and finally allocates the high-frequency component instructions to the supercapacitor, the low-frequency component instructions to the battery, and performs safety correction when the real-time data exceeds the preset safety threshold.

[0068] Specifically, the central management system, based on the system's long-term economic benefit goals, increases the upper limit of energy storage capacity from 1MWh to 2MWh and issues this strategy as part of the optimization objectives; the regional controller predicts that the charging station load distribution range is a peak power of 1MW±100kW, while the photovoltaic output uncertainty range is 800kW±50kW. The model predictive control algorithm calculates that the hybrid energy storage system will provide 200kW of power support in the next 15 minutes, and uses wavelet transform to decompose the 200kW power command, issuing the low-frequency component command and the high-frequency component command to the battery local controller and the supercapacitor local controller respectively; the battery local controller receives 1 Upon receiving a 50kW low-frequency component command, a dual closed-loop control of voltage and current is employed to smoothly discharge at a rate of 0.3C, avoiding losses caused by high-frequency fluctuations. By monitoring the battery's state of charge, the discharge power is automatically reduced to zero immediately when the real-time state of charge falls below a preset safety threshold of 20%, preventing over-discharge from damaging the battery. Upon receiving a 50kW high-frequency component command, the supercapacitor's local controller uses an inner current loop control to respond rapidly within milliseconds and absorb or release the power component, effectively smoothing out instantaneous power surges while also providing a safety correction function. When the real-time voltage exceeds 2.7V or the temperature exceeds 60℃, the charging and discharging current is automatically limited to ensure the safe operation of the supercapacitor.

[0069] The hierarchical control architecture proposed in this invention achieves seamless integration of macro-level strategies and micro-level execution through the collaborative work of a central management system, regional controllers, and local controllers. This architecture simplifies control logic, improves system scalability and reliability, and ensures accurate issuance and execution of instructions.

[0070] This invention effectively resolves the contradiction between high-frequency charging and discharging of charging stations and battery life by establishing a hybrid energy storage degradation cost function and incorporating this cost into a multi-objective model predictive control algorithm. It not only achieves short-term peak shaving and valley filling benefits, but also maximizes the long-term economic benefits and stability of the system throughout its entire life cycle.

[0071] Example 2:

[0072] The application scenario of this embodiment is when the charging station enters the off-peak electricity period at night. During this period, most vehicles are slowly charged, and the charging power is relatively stable. Occasionally, some electric vehicles will perform high-power fast charging, causing instantaneous spikes in the power exchange between the grid and the charging station.

[0073] As one embodiment of the present invention, refer to Figure 1 A flowchart of a hybrid energy storage method based on distributed photovoltaic power generation, referring to... Figure 2 This is a schematic diagram of a distributed control method based on distributed photovoltaic power generation and hybrid energy storage, referring to... Figure 3 This is a schematic diagram of the multi-objective model prediction algorithm structure.

[0074] Monitoring was conducted on an energy storage unit composed of lithium iron phosphate batteries. This unit has a capacity of 500 kWh and a unit capacity investment cost of 1000 yuan / kWh. During off-peak hours at night, most vehicles are slowly charging, resulting in a relatively stable load on the charging station. The batteries discharge at a 0.2C rate to meet load demands. Under this stable operation, the battery cycle life can reach over 4000 cycles, with a lifespan loss cost of 125 yuan per cycle. Occasionally, the demand for fast charging from electric vehicles can lead to high-power surges. For example, when the battery needs to discharge at a 1C rate to cope with instantaneous load spikes, its cycle life will decrease from 4000 cycles to 2000 cycles, with a lifespan loss cost of 250 yuan per cycle. During a high-power fast charging cycle at night, the supercapacitor needs to respond to more than 500 power pulses to smooth out high-frequency fluctuations during charging, averaging over 1000 power pulses. These pulse counts are converted into economic losses; for example, a loss of 5 yuan is set for every 1000 pulses, thus incorporating the losses from high-frequency charging and discharging into the total cost calculation.

[0075] Historical charging pile power demand data at charging stations during the night reveals an average nighttime load power of 500kW, but the instantaneous power can reach as high as 800kW when vehicles are fast charging. Simultaneously, user behavior data during this period, including vehicle access time, charging duration, and charging amount, is acquired. This data reveals the inherent patterns of nighttime load: the vast majority of vehicles primarily use slow charging, with charging durations of 6-8 hours and charging amounts of 30-50kWh between 22:00 and 00:00; a small number of fast-charging vehicles are randomly distributed throughout the night, typically charging for 30 minutes and charging for 20kWh. Feature engineering captures the periodicity and randomness of charging load, extracting periodic features such as hour and day of the week from timestamps. For example, the number of slow-charging vehicles is significantly higher on Friday and Saturday nights than on weekdays. Occasional fast-charging behaviors are labeled as random events and correlated with user behavior features such as vehicle access time and charging duration, which helps the model identify and predict potential high-power surges.

[0076] Based on historical data, the model predicts that between 02:00 and 02:15 the following day, due to the possibility of fast-charging vehicles accessing the system, the load power distribution range of the charging station will be 550kW to 1300kW. The model predictive control algorithm, based on stochastic optimization, calculates the optimal command: during off-peak hours, the battery is instructed to charge steadily at a 0.3C rate to store energy; during the 02:00-02:15 period, due to a predicted high power surge, the command is: the battery discharges steadily at a 0.5C rate to meet most of the slow-charging load demand; the supercapacitor is on standby, specifically designed to handle the predicted 550kW high-frequency power component to absorb the impact of sudden fast charging.

[0077] Between 02:00 and 02:15, the charging station experienced a power spike due to fast charging. The central management system calculated that the hybrid energy storage system needed to provide a power command containing the spike during this period. After wavelet transform decomposition of the command, the following was obtained: a low-frequency component of about 500kW to guide the battery to smoothly power most of the slow-charging vehicles; and a high-frequency component of about 300kW to guide the supercapacitor to respond quickly.

[0078] The battery local controller employs dual closed-loop control of voltage and current, smoothly discharging at approximately 500kW to handle low-frequency components and meet most slow-charging loads. Simultaneously, it continuously monitors the battery's real-time state of charge (SOC) and temperature. When the SOC falls below a preset safety threshold of 20%, the local controller automatically reduces the discharge power from 500kW to zero within 3 seconds, preventing irreversible damage from over-discharge. The supercapacitor local controller uses an inner-loop current control, enabling rapid response and handling of high-frequency components within milliseconds. When a fast-charging vehicle connects, causing a sudden increase in load, the supercapacitor local controller can quickly release approximately 300kW of power within 50 milliseconds, effectively absorbing the instantaneous power surge. It also performs safety corrections based on its real-time voltage and temperature; for example, if the voltage exceeds the safety threshold of 2.7V / cell or the temperature exceeds 60℃, it immediately limits the charging and discharging current to ensure equipment safety.

[0079] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A hybrid energy storage method based on distributed photovoltaic power generation, characterized in that, include: A battery degradation cost function is established based on the charge / discharge rate, power surge frequency, and pulse duration. A hybrid energy storage degradation cost function is established by combining the supercapacitor degradation cost function, and the long-term cost value is calculated. A multi-objective model predictive control algorithm is constructed to predict the uncertain range of photovoltaic output based on historical data and meteorological data, and to predict the load distribution range based on user behavior pattern data. Based on the long-term cost value, the uncertain range of photovoltaic output, and the load distribution range, the charging and discharging power command that minimizes the expected value of total operating cost and maximizes the utilization rate of photovoltaic power generation is calculated and decomposed into high-frequency and low-frequency components through wavelet transform. The real-time state of charge of the energy storage unit is obtained. Based on the real-time state of charge and the preset safety threshold, the local controller autonomously adjusts the charging and discharging power output, allocating high-frequency components to the supercapacitor and low-frequency components to the battery. Obtain operational data and the long-term cost value, construct a dynamic fitness function, and use the particle swarm optimization algorithm to solve for the optimal economic benefit of the photovoltaic and energy storage unit capacity configuration scheme; configure the capacity of the photovoltaic and energy storage units for the next cycle.

2. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, The implementation of the battery degradation cost function includes: establishing a mapping relationship between the actual cycle life of the battery and the battery health status under different discharge depths and charge / discharge rates based on historical operating data; quantifying the additional losses to the battery caused by the power surge during electric vehicle charging by using charge / discharge rate, power surge frequency, and pulse duration as variables; and calculating the economic losses to battery life caused by different charge / discharge behaviors by combining the battery's discharge depth, health status, and unit capacity investment cost. The specific implementation of the supercapacitor degradation cost function includes: collecting historical operating data and counting the actual number of cycles of the supercapacitor in charging and discharging scenarios; establishing a linear degradation model by correlating the number of cycles with the unit power investment cost of the supercapacitor, and calculating the economic losses of the supercapacitor in frequently coping with high-frequency fluctuations in charging load.

3. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, Historical photovoltaic power output data, meteorological data, and charging power demand data of charging piles are collected, as well as user behavior pattern data of charging stations. After preprocessing all collected data, feature engineering is performed to capture the periodicity and randomness of charging load. The uncertainty range of photovoltaic power output is predicted through a long short-term memory network, and the load distribution range of electricity load is predicted through a gradient boosting tree. The user behavior pattern data includes vehicle access time, charging duration, and charging amount.

4. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, A multi-objective optimization function is constructed using the predicted uncertain range of photovoltaic output, the load distribution range, and the degradation cost function as inputs. The power balance and state of charge safety thresholds of the energy storage unit are used as constraints. In each control cycle, stochastic optimization is performed based on the predicted uncertainty range to generate the expected total operating cost that minimizes the total operating cost and maximizes the photovoltaic power generation utilization rate. The minimized total operating cost includes the cost of purchasing and selling electricity, the degradation cost of hybrid energy storage, and the operation and maintenance cost.

5. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, The supercapacitor local controller uses current inner loop control to precisely adjust the charging and discharging current, quickly respond to and absorb high-frequency power components, and smooth out instantaneous power surges; the battery local controller uses voltage and current dual closed-loop control to smoothly handle low-frequency power components. Wavelet transform is used to process the charging and discharging power command output by the multi-objective model predictive control algorithm, decomposing the command into low-frequency and high-frequency components. Low-frequency components represent the long-term energy throughput requirements of the system and are used to guide the smooth charging and discharging of the battery. High-frequency components represent instantaneous power fluctuations in the system and are used to guide the rapid response of supercapacitors; The local controller autonomously adjusts its charging and discharging rate and power output, including: when the state of charge of the battery exceeds the preset safe operating range, the local controller of the battery automatically limits the charging and discharging power; The supercapacitor local controller limits the charging and discharging current when the voltage and temperature exceed a safety threshold, based on real-time voltage and temperature.

6. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, The dynamic fitness function balances the costs of electricity purchase and sale, maintenance costs, and hybrid energy storage degradation costs with peak shaving and valley filling benefits and grid ancillary service revenues, and calculates the maximum economic benefits using a particle swarm optimization algorithm. Based on the predicted load range, stochastic optimization is performed to output a capacity configuration scheme that extends battery life and meets the stochastic load demand of charging stations. Operational data includes historical aggregated state of charge data, energy utilization rate, battery health status, and grid interaction benefits.

7. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, The capacity configuration for the photovoltaic and energy storage units in the next cycle includes reserving redundant capacity for the future growth of electric vehicle charging stations in the capacity configuration scheme; according to the optimal capacity configuration scheme, the rated capacity of the photovoltaic power generation units and the rated capacity of the energy storage units are updated, and the updated capacity parameters are sent to the control logic of each layer controller as new operating constraints.

8. The method for hybrid energy storage based on distributed photovoltaic power generation according to claim 1, characterized in that, The central manager translates the capacity configuration scheme into macroscopic operating parameters and strategy instructions; the regional controllers receive the macroscopic operating parameters and strategy instructions, calculate the optimal charging and discharging power instructions through a multi-objective model predictive control algorithm, and decompose the power instructions into high-frequency and low-frequency components, issuing the component instructions as execution instructions; each energy storage unit and charging pile has a local controller that receives the component instructions, combines them with real-time operating data, and autonomously adjusts its charging and discharging rate and power output, allocating high-frequency component instructions to the supercapacitor and low-frequency component instructions to the battery, and performing safety corrections when real-time data exceeds a preset safety threshold.

Images