Flywheel lithium battery hybrid energy storage capacity configuration method and system

Abstract

The application discloses a flywheel lithium battery hybrid energy storage capacity configuration method and system, and particularly relates to the field of flywheel lithium batteries, and comprises the following steps: S1, collecting multi-dimensional operation parameters by adopting multi-sensor fusion, and performing data preprocessing by a Kalman filtering algorithm; S3, constructing a reinforcement learning collaborative model, obtaining equipment attenuation efficiency, and optimizing and outputting the optimal power distribution ratio of the flywheel and the lithium battery through model training; S5, constructing an attention mechanism model to predict future comprehensive power fluctuation, and simultaneously optimizing prediction accuracy; and S7, obtaining real-time actual capacity, setting double trigger conditions, updating the reinforcement learning model, and dynamically adjusting the capacity; through the multi-sensor fusion acquisition technology, the wavelet packet decomposition fluctuation extraction technology and the attention mechanism enhanced LSTM prediction technology, the application realizes accurate perception and prediction of new energy fluctuation and load demand, and solves the problem of low configuration accuracy caused by the dependence of the prior art on empirical formulas and static models.

Description

Technical Field

[0001] This invention relates to the field of flywheel lithium battery technology, and more specifically, to a method and system for configuring hybrid energy storage capacity of flywheel lithium batteries. Background Technology

[0002] With the large-scale application of new energy power generation technologies such as photovoltaics and wind power, the intermittency and fluctuation of their output power have a significant impact on the stable operation of the power grid. The flywheel energy storage and lithium battery hybrid energy storage system has become an important technical means to solve the problem of grid connection stability of new energy, thanks to the characteristics of flywheels such as fast response speed, long cycle life and no obvious limit on the number of charge and discharge cycles, and lithium batteries such as high energy density, large energy storage capacity and stable charge and discharge characteristics.

[0003] Existing flywheel-lithium battery hybrid energy storage capacity configurations, by combining the characteristics of both energy storage devices, alleviate the performance shortcomings of single energy storage systems to some extent, enabling complementary power and energy supply and improving the energy storage system's ability to mitigate fluctuations in new energy sources. However, significant shortcomings remain: First, capacity configurations largely rely on simplified empirical formulas or static models, failing to fully consider the dynamic changes in new energy output power and load demand, as well as the attenuation characteristics of energy storage devices during operation. They lack precise real-time data perception and dynamic adaptation technologies, leading to mismatches between configuration schemes and actual operational needs. Second, there is a lack of precise quantitative modeling and intelligent allocation strategies for the collaborative operation efficiency of flywheels and lithium batteries. Simply adding their capacities ignores the collaborative losses caused by differences in device response characteristics, reducing the overall system operating efficiency. Third, the entire lifecycle cost is not included as a core constraint in capacity configuration, and there is a lack of dynamic cost prediction methods based on digital twins. Focusing only on initial construction costs results in excessively high maintenance and loss costs during long-term operation, leading to poor economic efficiency. Fourth, there is a lack of dynamic adjustment mechanisms and edge computing support. Once the configuration scheme is determined, it cannot be optimized in real time based on device status and load changes, resulting in insufficient adaptability and flexibility.

[0004] To address the problems of low configuration accuracy, poor coordination efficiency, insufficient economy, and weak adaptability in existing technologies, this invention proposes a flywheel lithium battery hybrid energy storage capacity configuration method and system. Through precise acquisition of multi-dimensional parameters, intelligent extraction of fluctuation characteristics, dynamic optimization of coordination strategies, full life cycle cost simulation, enhanced load forecasting, intelligent capacity configuration, real-time dynamic adjustment, and virtual-real combined verification, the system achieves accurate, efficient, and economical capacity configuration of the hybrid energy storage system. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and system for configuring hybrid energy storage capacity using flywheel lithium batteries, which solves the problems mentioned in the background art through the following solutions.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for configuring the hybrid energy storage capacity of a flywheel lithium battery, comprising: S1. Multi-sensor fusion is used to collect multi-dimensional operating parameters, and data preprocessing is performed using the Kalman filter algorithm; S2. Wavelet packet decomposition is used to extract the energy entropy of each wave component, while determining the smoothing requirements and allocating the wave components according to frequency characteristics. S3. Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. S4. Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation; S5. Construct an attention mechanism model to predict future comprehensive power fluctuations, while optimizing prediction accuracy; S6. Pre-allocate initial capacity based on fluctuation component characteristics while performing cost verification, and determine the final initial capacity in combination with constraints. S7. Obtain the real-time actual capacity, set dual trigger conditions, update the reinforcement learning model, and dynamically adjust the capacity. S8. Build a virtual-real integrated verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration scheme.

[0007] Preferably, the multi-sensor fusion acquisition scheme is a combination of a laser power sensor, an ultrasonic condition monitoring sensor, and an electrochemical sensor, with a sampling frequency of 10Hz and a data transmission delay of ≤50ms. The multi-dimensional operating parameters include new energy side parameters, load side parameters, flywheel equipment parameters, lithium battery equipment parameters, and system constraint parameters. The new energy side parameters specifically include the instantaneous value of photovoltaic output power. Instantaneous value of wind power output The load-side parameters specifically include real-time power demand. Power fluctuation allowable threshold The specific parameters of the flywheel device include the maximum charging and discharging power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific parameters of lithium battery equipment include maximum charge and discharge power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific system constraint parameters include the total lifespan T of the energy storage system, the discount rate r, and the capacity decay threshold. Prediction accuracy threshold System overall efficiency threshold Lifecycle cost threshold .

[0008] Preferably, the wavelet packet decomposition uses db4 basis functions to respectively decompose the wavelet packets. A three-level decomposition is performed to separate high-frequency instantaneous fluctuations from low-frequency continuous fluctuations; the energy entropy of the fluctuation components includes the energy entropy of the photovoltaic fluctuation components. Wind power fluctuation component energy entropy and load fluctuation component energy entropy ,in This represents the energy percentage of the i-th component of the photovoltaic fluctuation. This represents the energy percentage of the i-th component of wind power fluctuations. This represents the energy percentage of the i-th component of the load fluctuation; the demand mitigation is achieved through the comprehensive power fluctuation value. and power exceeding the limit Confirmed; the fluctuation component allocation rule is that high-frequency fluctuation components are allocated to the flywheel response, and low-frequency fluctuation components are allocated to the lithium battery response.

[0009] Preferably, the reinforcement learning collaborative model includes a state space definition: Motion space definition: flywheel output power percentage Lithium battery output power ratio Reward function definition: ,in For collaborative efficiency, The current cost; the device degradation efficiency includes the actual charge / discharge efficiency after flywheel degradation. And the actual charge / discharge efficiency after lithium battery degradation The model training employs the DQN algorithm, with ≥5000 iterations and a convergence error ≤1%; the optimal power allocation ratio... Determine the actual output power of the flywheel after output. Actual output power of lithium battery And it satisfies the power constraint conditions.

[0010] Preferably, digital twin technology constructs a digital twin of the flywheel-lithium battery hybrid energy storage system to map the physical attributes, operating status, and environmental influencing factors of the equipment in real time, and synchronizes equipment degradation data and operating loss data; the full life cycle cost model ,in Indicates the flywheel's energy storage capacity. Indicates the energy storage capacity of lithium batteries. This indicates the predicted decay efficiency of the flywheel at different times. This indicates the predicted degradation efficiency of the lithium battery at different times. Let r represent the total lifecycle cost, r represent the discount rate, and t represent the time integration variable; the constraints include power constraints. Capacity constraints Smoothing constraints Lifetime constraints and content decay throughout the entire lifecycle .

[0011] Preferably, the attention mechanism model uses an enhanced LSTM model to divide historical data into training and testing sets in a 7:3 ratio, with the input sequence length set to 24. The attention mechanism layer uses the formula... (in Assigning weights to features at different times in the input sequence, the model predicts future time windows. Comprehensive power fluctuation within Using mean absolute error Evaluation accuracy, among which This represents the input features at time t in the attention mechanism. 'b' represents the training parameters for the attention mechanism. This represents the feature score at time t in the attention mechanism. This indicates the predicted overall power fluctuation within a future time window. This represents the actual value of the comprehensive power fluctuation; the optimized prediction accuracy is adjusted using an adaptive learning rate, with an initial learning rate of 0.001, decreasing by 10% every 100 iterations until... .

[0012] Preferably, the pre-allocated initial capacity includes the initial pre-allocated capacity of the flywheel. Initial pre-allocated capacity of lithium batteries The cost verification is performed by calculating the pre-allocated capacity using a digital twin cost model. ,like Then adjust according to the coefficient. Adjust the pre-allocated capacity; the method for determining the final initial capacity is as follows: substitute the adjusted pre-allocated capacity into the constraints in S4; if any constraint is not met, then re-optimize the power allocation ratio through a reinforcement learning model. Repeat steps S1-S2 until all constraints are satisfied, then output the final initial capacity: ; ; in The final initial capacity of the flywheel, This represents the final initial capacity of the lithium battery.

[0013] Preferably, the method for obtaining the real-time actual capacity is as follows: Real-time operating data of the device is collected every 5 minutes through a dynamic adjustment mechanism driven by edge computing to obtain the real-time actual capacity. The triggering condition is: or ,and ; Re-optimize by updating the state variables of the reinforcement learning model Adjust the formula to .

[0014] Preferably, the virtual-real hybrid verification platform includes a physical end and a virtual end. The physical end builds a small-scale hybrid energy storage experimental platform and connects to a simulated new energy power generation system and load system; the virtual end calls a digital twin to synchronize the operating data of the physical end; the overall system efficiency... The total lifecycle cost The method for verifying the validity of the configuration scheme is as follows: Furthermore, the performance error between the physical and virtual ends must be ≤3%; otherwise, readjustment will be performed.

[0015] Preferably, a flywheel lithium battery hybrid energy storage capacity configuration system includes: Initial module: Multi-sensor fusion is used to collect multi-dimensional operating parameters, and the data is preprocessed using the Kalman filter algorithm; The extraction and allocation module uses wavelet packet decomposition to extract the energy entropy of each fluctuation component, while determining the smoothing demand and allocating the fluctuation components according to frequency characteristics. Reinforcement learning module: Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. Cost Constraint Module: Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation; Attention Prediction Module: Constructs an attention mechanism model to predict future overall power fluctuations while optimizing prediction accuracy; Pre-allocation verification module: Based on the characteristics of fluctuation components, the initial capacity is pre-allocated and cost verification is performed simultaneously, and the final initial capacity is determined in combination with constraints. Edge computing module: acquires real-time actual capacity, sets dual trigger conditions, updates reinforcement learning models, and dynamically adjusts capacity; Solution verification module: Build a virtual-physical verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration solution.

[0016] The technical effects and advantages of this invention are as follows: 1. This invention achieves accurate perception and prediction of new energy fluctuations and load demand through multi-sensor fusion acquisition technology, wavelet packet decomposition fluctuation extraction technology, and LSTM prediction technology enhanced by attention mechanism. It solves the problem of low configuration accuracy caused by existing technologies relying on empirical formulas and static models, and makes the configuration of hybrid energy storage capacity highly matched with actual operation requirements, significantly improving the effect of new energy fluctuation smoothing and ensuring the stability of power grid operation. 2. This invention introduces reinforcement learning collaborative strategy and digital twin simulation technology to construct a dynamic power allocation model and full life cycle cost simulation constraints, which solves the problems of poor collaborative efficiency and insufficient economic efficiency of existing technologies. While realizing efficient collaborative operation of flywheel and lithium battery, it accurately controls the full life cycle cost, improves the economic feasibility and sustainability of the project, and extends the actual service life of the equipment. 3. This invention builds a real-time dynamic adjustment mechanism based on the edge computing processing module, and ensures the effectiveness of the solution through a virtual-real combined optimization verification platform. It solves the problems of weak adaptability and poor implementation of existing technologies and configuration schemes, realizes real-time optimization and dynamic adaptation of capacity configuration, improves the adaptability of hybrid energy storage systems to complex operating conditions, and provides innovative and practical technical support for the engineering application of hybrid energy storage systems. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0018] 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.

[0019] As attached Figure 1 The method for configuring hybrid energy storage capacity using a flywheel lithium battery, as shown, includes the following steps: S1. Multi-sensor fusion is used to collect multi-dimensional operating parameters, and data preprocessing is performed using the Kalman filter algorithm; Specifically, it should be noted that the multi-sensor fusion acquisition scheme is a combination of a laser power sensor, an ultrasonic condition monitoring sensor, and an electrochemical sensor, with a sampling frequency of 10Hz and a data transmission delay of ≤50ms; the multi-dimensional operating parameters include new energy side parameters, load side parameters, flywheel equipment parameters, lithium battery equipment parameters, and system constraint parameters; the new energy side parameters specifically include the instantaneous value of photovoltaic output power. Instantaneous value of wind power output The load-side parameters specifically include real-time power demand. Power fluctuation allowable threshold The specific parameters of the flywheel device include the maximum charging and discharging power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific parameters of lithium battery equipment include maximum charge and discharge power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific system constraint parameters include the total lifespan T of the energy storage system, the discount rate r, and the capacity decay threshold. Prediction accuracy threshold System overall efficiency threshold Lifecycle cost threshold Specifically, the core prerequisite for capacity configuration is mastering comprehensive information on renewable energy output, load demand, energy storage equipment performance, and cost. Existing technologies suffer from limitations in data acquisition and accuracy, while multi-sensor fusion can overcome the measurement limitations of single sensors, ensuring data coverage and accuracy. This provides a reliable foundation for all subsequent analysis and calculations, aligning with the technical logic of "data-driven decision-making." Laser power sensors, ultrasonic condition monitoring sensors, and electrochemical sensors are all mature industrial measurement devices. Kalman filtering noise reduction algorithms are classic data preprocessing techniques, and the communication stability of industrial Ethernet ensures data transmission across multiple modules. The entire acquisition solution has a clear technical implementation path and no technical barriers. By collaboratively acquiring and preprocessing data through multiple sensors, the problem of insufficient data support in existing technologies is solved. This ensures that subsequent steps such as fluctuation extraction, collaborative modeling, and cost calculation are based on real and reliable parameters, improving the accuracy of capacity configuration from the source and laying a data foundation for the feasibility of the entire technical solution.

[0020] S2. Wavelet packet decomposition is used to extract the energy entropy of each wave component, while determining the smoothing requirements and allocating the wave components according to frequency characteristics. Specifically, it should be noted that the wavelet packet decomposition uses db4 basis functions to respectively... A three-level decomposition is performed to separate high-frequency instantaneous fluctuations from low-frequency continuous fluctuations; the energy entropy of the fluctuation components includes the energy entropy of the photovoltaic fluctuation components. Wind power fluctuation component energy entropy and load fluctuation component energy entropy ,in This represents the energy percentage of the i-th component of the photovoltaic fluctuation. This represents the energy percentage of the i-th component of wind power fluctuations. This represents the energy percentage of the i-th component of the load fluctuation; the demand mitigation is achieved through the comprehensive power fluctuation value. and power exceeding the limit It is confirmed that the fluctuation component allocation rule is to allocate high-frequency fluctuation components to the flywheel response and low-frequency fluctuation components to the lithium battery response. Specifically, the rationale is as follows: fluctuations in new energy sources and loads include high-frequency instantaneous fluctuations and low-frequency continuous fluctuations. The response characteristics of flywheels and lithium batteries differ (flywheels are suitable for rapid responses to high-frequency fluctuations, while lithium batteries are suitable for stable responses to low-frequency fluctuations). Simple calculations alone cannot accurately distinguish the fluctuation types. Wavelet packet decomposition can effectively separate fluctuation components of different frequencies, providing a basis for targeted allocation of energy storage tasks. Wavelet packet decomposition technology has been widely used in signal processing for fluctuation feature extraction. The db4 basis function is a mature decomposition function that has been verified in engineering. Energy entropy calculation can quantify the complexity of fluctuations. The entire technical method has complete theoretical support and engineering application cases, and can be directly implemented. This solves the problem of ambiguous fluctuation feature identification in existing technologies. By accurately separating fluctuation components and clearly identifying the excess power that needs to be smoothed, the capacity configuration of flywheels and lithium batteries becomes more targeted, avoiding blindly configuring "large and comprehensive" components. This provides key support for improving the smoothing efficiency of energy storage systems and reducing redundant capacity.

[0021] S3. Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. Specifically, it should be noted that the reinforcement learning collaborative model includes a state space definition: Motion space definition: flywheel output power percentage Lithium battery output power ratio Reward function definition: ,in For collaborative efficiency, The current cost; the device degradation efficiency includes the actual charge / discharge efficiency after flywheel degradation. And the actual charge / discharge efficiency after lithium battery degradation The model training employs the DQN algorithm, with ≥5000 iterations and a convergence error ≤1%; the optimal power allocation ratio... Determine the actual output power of the flywheel after output. Actual output power of lithium battery Furthermore, it satisfies power constraints. Specifically, flywheels and lithium batteries differ in response time, charge / discharge efficiency, and degradation characteristics. Simple fixed-ratio allocation leads to collaborative losses, while reinforcement learning can find the optimal power allocation solution through dynamic training. This adapts to fluctuating demand, maximizes collaborative efficiency, and minimizes costs, aligning with the technical goal of "dynamically optimized collaboration." The DQN algorithm in reinforcement learning is a mature deep reinforcement learning framework that has been applied in energy system optimization and equipment collaborative control. Its state space and action space definitions conform to the actual scenario of hybrid energy storage collaboration, and the iterative training mechanism ensures that the model converges to the optimal solution, validating its technical feasibility. It overcomes the limitations of existing technologies that rely on "simple capacity superposition" for collaboration, achieving complementary advantages through intelligent power allocation, solving the problem of low collaborative efficiency, and providing accurate power allocation basis for subsequent capacity calculations. This ensures a high degree of matching between capacity configuration and collaborative strategy, improving the overall system operating efficiency.

[0022] S4. Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation; Specifically, it should be noted that: digital twin technology, by building a digital twin of the flywheel-lithium battery hybrid energy storage system, maps the physical attributes, operating status, and environmental influencing factors of the equipment in real time, and synchronizes equipment degradation data and operating loss data; the aforementioned life cycle cost model ,in Indicates the flywheel's energy storage capacity. Indicates the energy storage capacity of lithium batteries. This indicates the predicted decay efficiency of the flywheel at different times. This indicates the predicted degradation efficiency of the lithium battery at different times. Let r represent the total lifecycle cost, r represent the discount rate, and t represent the time integration variable; the constraints include power constraints. Capacity constraints Smoothing constraints Lifetime constraints and content decay throughout the entire lifecycle Specifically, existing technologies only focus on initial construction costs, neglecting long-term expenditures such as operating losses and maintenance costs due to equipment degradation. Digital twins, however, can simulate the entire lifecycle of equipment through virtual-real mapping, accurately predict degradation trends, and thus construct a comprehensive cost model, ensuring that capacity configuration meets both technical and economic requirements. Digital twin technology has already been implemented in industrial fields such as energy storage and power. It can construct a virtual twin of an energy storage system through 3D modeling and real-time data synchronization. The composition of the entire lifecycle cost (construction cost, operating loss cost, and maintenance cost) conforms to the logic of engineering economic accounting, and the discount rate calculation reflects the time value of money. The technical solution has a mature implementation path. It solves the problem of one-sided economic considerations in existing technologies by using digital twin simulation to achieve dynamic cost prediction and multi-constraint verification, ensuring that the capacity configuration scheme is cost-optimal throughout its entire lifecycle. This avoids project unsustainability due to excessive costs in the later stages, thus improving the economic feasibility of the technical solution.

[0023] S5. Construct an attention mechanism model to predict future comprehensive power fluctuations, while optimizing prediction accuracy; Specifically, it should be noted that the attention mechanism model uses an enhanced LSTM model to divide historical data into training and testing sets in a 7:3 ratio, with the input sequence length set to 24. The attention mechanism layer uses the formula... (in Assigning weights to features at different times in the input sequence, the model predicts future time windows. Comprehensive power fluctuation within Using mean absolute error Evaluation accuracy, among which This represents the input features at time t in the attention mechanism. 'b' represents the training parameters for the attention mechanism. This represents the feature score at time t in the attention mechanism. This indicates the predicted overall power fluctuation within a future time window. This represents the actual value of the comprehensive power fluctuation; the optimized prediction accuracy is adjusted using an adaptive learning rate, with an initial learning rate of 0.001, decreasing by 10% every 100 iterations until... Specifically, regarding rationality: capacity allocation needs to consider both current fluctuating demands and future trends. Existing forecasting models ignore the impact of data from critical periods, leading to insufficient prediction accuracy. The LSTM algorithm, however, excels at processing time-series data, and its attention mechanism strengthens the weights of features from important periods, improving forecast targeting and meeting the technical requirement of "forward-looking allocation." The LSTM algorithm is a classic model in the field of time-series forecasting. The fusion technology of the attention mechanism has been widely applied in scenarios such as load forecasting and new energy output forecasting. Adaptive learning rate adjustment ensures model convergence, and parameter settings such as the ratio of training and test sets and the length of input sequences conform to engineering practice, demonstrating high technical maturity. This solves the problem of insufficient prediction accuracy in existing technologies. By accurately predicting future fluctuating demands, capacity allocation can adapt to changes in advance, avoiding under-allocation or redundancy due to prediction bias. It provides a forward-looking basis for initial capacity allocation and dynamic adjustments, improving the adaptability and flexibility of capacity allocation.

[0024] S6. Pre-allocate initial capacity based on fluctuation component characteristics while performing cost verification, and determine the final initial capacity in combination with constraints. Specifically, it should be noted that the pre-allocated initial capacity includes the initial pre-allocated capacity of the flywheel. Initial pre-allocated capacity of lithium batteries The cost verification is performed by calculating the pre-allocated capacity using a digital twin cost model. ,like Then adjust according to the coefficient. Adjust the pre-allocated capacity; the method for determining the final initial capacity is as follows: substitute the adjusted pre-allocated capacity into the constraints in S4; if any constraint is not met, then re-optimize the power allocation ratio through a reinforcement learning model. Repeat steps S1-S2 until all constraints are satisfied, then output the final initial capacity: ; ; in The final initial capacity of the flywheel, This refers to the final initial capacity of the lithium battery. Specifically, regarding rationality: direct calculation of initial capacity configuration can easily lead to non-compliance with constraints. The closed-loop mechanism of "pre-allocation-verification feedback" can initially allocate capacity based on fluctuation characteristics and collaborative strategies, and then verify and adjust it through cost and technical constraints. This ensures the correct configuration direction while avoiding violations, conforming to the engineering logic of "step-by-step optimization." Regarding existence: pre-allocation is based on the calculation logic of equipment response characteristics and collaborative efficiency. The verification feedback mechanism is a commonly used optimization method in engineering design. The calculation of adjustment coefficients can balance cost and demand. The entire process does not rely on special technologies; it can be achieved solely through existing calculation logic and constraint judgments, with clear operational steps. Significance: This solves the problem of blind initial configuration in existing technologies. Through closed-loop optimization, it outputs an initial capacity that meets technical, economic, and constraint conditions, avoiding obvious defects in the initial scheme and providing a stable benchmark for subsequent dynamic adjustments, ensuring that the capacity configuration is feasible and optimal from the initial stage.

[0025] S7. Obtain the real-time actual capacity, set dual trigger conditions, update the reinforcement learning model, and dynamically adjust the capacity. Specifically, it should be noted that the method for obtaining the real-time actual capacity is as follows: Real-time operating data of the device is collected every 5 minutes through a dynamic adjustment mechanism driven by edge computing to obtain the real-time actual capacity. The triggering condition is: or ,and ; Re-optimize by updating the state variables of the reinforcement learning model Adjust the formula to Specifically, energy storage devices degrade over time, and load and renewable energy fluctuations change in real time. Fixed configuration schemes cannot adapt to dynamic changes, while edge computing, with its low latency and real-time data processing capabilities, can quickly monitor device status and fluctuations, triggering adjustment mechanisms and meeting the technical requirement of "real-time adaptation." Edge computing modules are widely used in industrial real-time control scenarios. Their data processing speed and low latency characteristics can meet the monitoring frequency requirement of once every 5 minutes. The real-time update mechanism of the reinforcement learning model has theoretical support, and the derivation of the dynamic adjustment formula conforms to the matching logic of capacity with fluctuations and degradation efficiency, making the technology highly feasible. This solves the problem of the lack of dynamic adjustment mechanisms in existing technologies. By processing data and optimizing capacity configuration in real time through edge computing, it ensures that the energy storage system can maintain optimal operating status even under device degradation and fluctuations, avoiding a disconnect between the configuration scheme and actual needs, and improving the system's adaptability to complex operating conditions.

[0026] S8. Build a virtual-real integrated verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration scheme.

[0027] Specifically, it should be noted that the virtual-real hybrid verification platform includes a physical end and a virtual end. The physical end builds a small-scale hybrid energy storage experimental platform, connecting to a simulated new energy power generation system and load system; the virtual end calls a digital twin to synchronize the operating data of the physical end; the overall efficiency of the system... The total lifecycle cost The method for verifying the validity of the configuration scheme is as follows: Furthermore, the performance error between the physical and virtual ends must be ≤3%; otherwise, readjustment is required. Specifically, regarding rationality: theoretically calculated configuration schemes may be at risk of deviating from actual operation. Virtual simulation alone cannot verify the actual performance of physical equipment, and physical experiments alone cannot cover various operating conditions. Combining virtual and physical methods balances the efficiency of simulation and the authenticity of experiments, ensuring the scheme meets requirements in both theory and practice. The small flywheel-lithium battery hybrid energy storage experimental platform can be built using existing equipment. Digital twins can synchronize physical data for real-time simulation. Performance error comparison is a common method for engineering verification. The entire verification scheme combines mature experience in simulation and experimental technologies and has a clear implementation process. It solves the problem of poor feasibility of existing technical configuration schemes. Verification through a combination of virtual and physical methods ensures the effectiveness and reliability of the scheme, avoiding discrepancies between theoretical design and actual operation. It also provides experimental support for the finalization of the scheme, ensuring that it meets the requirements of overall system efficiency and total lifecycle cost in engineering applications, guaranteeing the practical value of the technical solution.

[0028] Based on the above scheme and appendix Figure 2 The present invention also discloses a flywheel lithium battery hybrid energy storage capacity configuration system, comprising: Initial module: Multi-sensor fusion is used to collect multi-dimensional operating parameters, and the data is preprocessed using the Kalman filter algorithm; The extraction and allocation module uses wavelet packet decomposition to extract the energy entropy of each fluctuation component, while determining the smoothing demand and allocating the fluctuation components according to frequency characteristics. Reinforcement learning module: Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. Cost Constraint Module: Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation; Attention Prediction Module: Constructs an attention mechanism model to predict future overall power fluctuations while optimizing prediction accuracy; Pre-allocation verification module: Based on the characteristics of fluctuation components, the initial capacity is pre-allocated and cost verification is performed simultaneously, and the final initial capacity is determined in combination with constraints. Edge computing module: acquires real-time actual capacity, sets dual trigger conditions, updates reinforcement learning models, and dynamically adjusts capacity; Solution verification module: Build a virtual-physical verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration solution.

[0029] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for configuring the hybrid energy storage capacity of a flywheel lithium battery, characterized in that, include: S1. Multi-sensor fusion is used to collect multi-dimensional operating parameters, and data preprocessing is performed using the Kalman filter algorithm; S2. Wavelet packet decomposition is used to extract the energy entropy of each wave component, while determining the smoothing requirements and allocating the wave components according to frequency characteristics. S3. Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. The reinforcement learning collaborative model includes a state space definition: power excess portion Flywheel charging and discharging efficiency Lithium battery charge and discharge efficiency Flywheel response time Lithium battery response time Motion space definition: flywheel output power percentage Lithium battery output power ratio Reward function definition: ,in For collaborative efficiency, At current cost, The threshold for total lifecycle cost; The device attenuation efficiency includes the actual charge / discharge efficiency after flywheel attenuation. And the actual charge / discharge efficiency after lithium battery degradation The model training employs the DQN algorithm, with ≥5000 iterations and a convergence error ≤1%; the optimal power allocation ratio... Determine the actual output power of the flywheel after output. Actual output power of lithium battery And it satisfies the power constraint condition; S4. Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation. Digital twin technology constructs a digital twin of a flywheel-lithium battery hybrid energy storage system, mapping the physical attributes, operating status, and environmental influencing factors of the equipment in real time, and synchronizing equipment degradation data and operating loss data; the aforementioned life-cycle cost model ,in This represents the construction cost per unit capacity. Indicates the flywheel's energy storage capacity. This represents the construction cost per unit capacity. This represents the annual operating loss cost per unit capacity of the flywheel. Indicates the energy storage capacity of lithium batteries. This indicates the predicted decay efficiency of the flywheel at different times. This indicates the annual operating loss cost per unit capacity of a lithium battery. This indicates the predicted degradation efficiency of the lithium battery at different times. This indicates the annual maintenance cost per unit capacity of the flywheel. The value represents the annual maintenance cost per unit capacity of a lithium battery, r represents the discount rate, and t represents the time integral variable; the constraints include power constraints. Capacity constraints Smoothing constraints Lifetime constraints and content decay throughout the entire lifecycle ,in This represents the maximum charging and discharging power of the flywheel. This represents the maximum charge and discharge power of the lithium battery. The continuous running time of the flywheel. This refers to the continuous operating time of the lithium battery. S5. Construct an attention mechanism model to predict future comprehensive power fluctuations, while optimizing prediction accuracy; S6. Pre-allocate initial capacity based on fluctuation component characteristics while performing cost verification, and determine the final initial capacity in combination with constraints. The pre-allocated initial capacity includes the initial pre-allocated capacity of the flywheel. Initial pre-allocated capacity of lithium batteries ; The cost verification calculates the pre-allocated capacity using a digital twin cost model. ,like Then adjust according to the coefficient. Adjust the pre-allocated capacity; the method for determining the final initial capacity is as follows: substitute the adjusted pre-allocated capacity into the constraints in S4; if any constraint is not met, then re-optimize the power allocation ratio through a reinforcement learning model. Repeat steps S1-S2 until all constraints are satisfied, then output the final initial capacity: ； ； in The final initial capacity of the flywheel, This represents the final initial capacity of the lithium battery. S7. Obtain the real-time actual capacity, set dual trigger conditions, update the reinforcement learning model, and dynamically adjust the capacity. S8. Build a virtual-real integrated verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration scheme.

2. The method for configuring the hybrid energy storage capacity of a flywheel lithium battery according to claim 1, characterized in that: The multidimensional operating parameters include renewable energy side parameters, load side parameters, flywheel equipment parameters, lithium battery equipment parameters, and system constraint parameters; the renewable energy side parameters specifically include the instantaneous value of photovoltaic output power. Instantaneous value of wind power output The load-side parameters specifically include real-time power demand. Power fluctuation allowable threshold The specific parameters of the flywheel device include the maximum charging and discharging power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific parameters of lithium battery equipment include maximum charge and discharge power. Charge and discharge efficiency Response time Total number of cycles Number of loops used Unit capacity construction cost Annual operating loss cost per unit capacity Annual maintenance cost per unit capacity attenuation coefficient Continuous running time The specific system constraint parameters include the total lifespan T of the energy storage system, the discount rate r, and the capacity decay threshold. Prediction accuracy threshold System overall efficiency threshold Lifecycle cost threshold .

3. The method for configuring the hybrid energy storage capacity of a flywheel lithium battery according to claim 2, characterized in that: The wavelet packet decomposition uses db4 basis functions to respectively... A three-level decomposition is performed to separate high-frequency instantaneous fluctuations from low-frequency continuous fluctuations; the energy entropy of the fluctuation components includes the energy entropy of the photovoltaic fluctuation components. Wind power fluctuation component energy entropy and load fluctuation component energy entropy ,in This represents the energy percentage of the i-th component of the photovoltaic fluctuation. This represents the energy percentage of the i-th component of wind power fluctuations. This represents the energy percentage of the i-th component of the load fluctuation; the demand mitigation is achieved through the comprehensive power fluctuation value. and power exceeding the limit Confirmed; the fluctuation component allocation rule is that high-frequency fluctuation components are allocated to the flywheel response, and low-frequency fluctuation components are allocated to the lithium battery response.

4. The method for configuring the hybrid energy storage capacity of a flywheel lithium battery according to claim 2, characterized in that: The attention mechanism model uses an enhanced LSTM model to divide historical data into training and testing sets in a 7:3 ratio. The input sequence length is set to 24. The attention mechanism layer uses the formula... The model assigns feature weights to the input sequence at different times, and then predicts future time windows. Comprehensive power fluctuation within Using mean absolute error Evaluation accuracy, among which , This represents the input features at time t in the attention mechanism. 'b' represents the training parameters for the attention mechanism. This represents the feature score at time t in the attention mechanism. This represents the predicted overall power fluctuation within a future time window. This represents the actual value of the comprehensive power fluctuation; the optimized prediction accuracy is adjusted using an adaptive learning rate, with an initial learning rate of 0.001, decreasing by 10% every 100 iterations until... .

5. The method for configuring the hybrid energy storage capacity of a flywheel lithium battery according to claim 1, characterized in that: The method for obtaining the real-time actual capacity is as follows: Real-time operating data of the device is collected every 5 minutes through a dynamic adjustment mechanism driven by edge computing to obtain the real-time actual capacity of the flywheel. Real-time actual capacity of lithium batteries The triggering condition is: or ,and ; Re-optimize by updating the state variables of the reinforcement learning model The adjusted capacity of the lithium battery is obtained after dynamic adjustment. ； ； in This indicates the flywheel attenuation efficiency in the initial configuration. This indicates the lithium battery degradation efficiency at the initial configuration. This indicates the flywheel's real-time decay efficiency. This indicates the real-time degradation efficiency of the lithium battery.

6. The method for configuring the hybrid energy storage capacity of a flywheel lithium battery according to claim 5, characterized in that: The virtual-physical hybrid verification platform includes a physical terminal and a virtual terminal. The physical terminal builds a small-scale hybrid energy storage experimental platform, which is connected to a simulated new energy power generation system and a load system. The virtual terminal calls a digital twin to synchronize the operating data of the physical terminal. The method for verifying the effectiveness of the configuration scheme is as follows: Furthermore, the performance error between the physical and virtual ends must be ≤3%; otherwise, readjustment will be performed.

7. A flywheel lithium battery hybrid energy storage capacity configuration system, used to implement the flywheel lithium battery hybrid energy storage capacity configuration method according to any one of claims 1-6, characterized in that, include: Initial module: Multi-sensor fusion is used to collect multi-dimensional operating parameters, and the data is preprocessed using the Kalman filter algorithm; The extraction and allocation module uses wavelet packet decomposition to extract the energy entropy of each fluctuation component, while determining the smoothing demand and allocating the fluctuation components according to frequency characteristics. Reinforcement learning module: Construct a reinforcement learning collaborative model and obtain the device attenuation efficiency. Through model training, optimize and output the optimal power allocation ratio between the flywheel and the lithium battery. Cost Constraint Module: Build a digital twin of the hybrid energy storage system, construct a full life cycle cost model, set constraints, and conduct cost simulation; Attention Prediction Module: Constructs an attention mechanism model to predict future overall power fluctuations while optimizing prediction accuracy; Pre-allocation verification module: Based on the characteristics of fluctuation components, the initial capacity is pre-allocated and cost verification is performed simultaneously, and the final initial capacity is determined in combination with constraints. Edge computing module: Obtains real-time actual capacity, sets dual trigger conditions, updates reinforcement learning models, and dynamically adjusts capacity; Solution verification module: Build a virtual-physical verification platform, run data synchronously, calculate the overall system efficiency and lifecycle cost, and verify the effectiveness of the configuration solution.

Images