A method, system and electronic device for energy storage system control for photovoltaic assemblies

Abstract

The application provides a method, system and electronic device for controlling an energy storage system of a photovoltaic assembly, the method comprising: collecting real-time output power signals and related operating environment data of the photovoltaic assembly; performing empirical mode decomposition on the output power signals to obtain a plurality of intrinsic mode function components and a residual; performing Hilbert transform on each intrinsic mode function component to extract the instantaneous characteristics of each intrinsic mode function component, the instantaneous characteristics including instantaneous frequency and instantaneous amplitude; synthesizing a photovoltaic power prediction value at a future time according to the instantaneous characteristics of the intrinsic mode function components and the trend prediction of the residual; constructing a multi-objective function model according to the photovoltaic power prediction value; solving the multi-objective optimization model under the condition of satisfying power balance constraints and energy storage unit state of charge constraints to obtain a power distribution instruction; and controlling the energy storage system to charge and discharge according to the power distribution instruction.

Description

Technical Field

[0001] This application relates to the field of photovoltaic energy storage technology, and in particular to a control method, system and electronic equipment for an energy storage system used in photovoltaic modules. Background Technology

[0002] As the global energy structure accelerates its transition towards cleaner and lower-carbon energy, photovoltaic (PV) power generation has become a significant growth driver due to its fuel-free and renewable nature. However, as the proportion of PV installations increases, its inherent intermittency and rapid time-varying characteristics pose new challenges to grid stability and system economics. Changes in sunlight can cause a sudden drop in array output within seconds, while the grid's frequency and voltage regulation requirements often require millisecond-level responses. Traditional centralized energy storage and control systems often struggle to track and suppress such second-level disturbances due to communication delays and the limited ramp-up rates of energy storage devices, leading to risks such as frequency deviations, harmonics, or protection malfunctions. At the same time, frequent charging and discharging accelerates lithium battery capacity degradation, driving up PV costs and impairing long-term economic viability.

[0003] In related technologies, existing solutions lack a unified multi-objective optimization framework, and existing communication and coordination protocols, while ensuring rapid response and reliability, often face limitations in computational resources and insufficient solution speed. Based on these shortcomings, there is an urgent need for a systematic solution encompassing physical integration design, second-level prediction, and hierarchical hybrid energy storage coordinated control to achieve safe, economical, and efficient grid connection for high-penetration photovoltaic systems. Summary of the Invention

[0004] In view of this, it is necessary to provide a control method, system and electronic equipment for energy storage systems of photovoltaic modules that can overcome at least one of the above defects.

[0005] In a first aspect, embodiments of this application provide a method for controlling an energy storage system for photovoltaic modules, applied to controlling photovoltaic modules with an energy storage system, the method comprising: Collect the real-time output power signal and related operating environment data of the photovoltaic module; Empirical mode decomposition is performed on the output power signal to obtain several intrinsic mode function components and residuals; A Hilbert transform is performed on each of the intrinsic mode function components to extract the instantaneous features of each intrinsic mode function component, the instantaneous features including instantaneous frequency and instantaneous amplitude; Based on the instantaneous characteristics of the intrinsic mode function components and the trend prediction of the residuals, the predicted photovoltaic power value for future times is synthesized. A multi-objective function model is constructed based on the photovoltaic power prediction value. The multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life and system operating cost of photovoltaic modules. Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, the multi-objective optimization model is solved to obtain power allocation instructions; According to the power allocation command, the energy storage system is controlled to perform coordinated charging and discharging.

[0006] In one embodiment, the empirical mode decomposition of the output power signal includes: Identify all local extrema of the output power signal; Using the aforementioned local extreme points, the upper and lower envelopes are constructed using cubic spline interpolation. Calculate the envelope mean of the upper envelope and the lower envelope; Subtract the envelope mean from the output power signal to obtain the first undetermined intrinsic mode function component; Determine whether the first undetermined intrinsic mode function component satisfies the conditions for intrinsic mode functions; If satisfied, the first undetermined intrinsic mode function component is recorded as the first intrinsic mode function component; The first intrinsic mode function component is separated from the output power signal to obtain the first residual term; The residual term is iterated repeatedly to form a new output power signal until the final residual term is a monotonic function or a constant, thereby obtaining at least one intrinsic mode function component and the residual term.

[0007] In one embodiment, performing a Hilbert transform on each of the intrinsic mode function components includes: A Hilbert transform is performed on each of the intrinsic mode function components to obtain the imaginary part of the corresponding analytic signal; Based on the analytical signal, the instantaneous amplitude, instantaneous phase, and instantaneous angular frequency corresponding to each of the intrinsic mode function components are constructed; The instantaneous amplitude is used to describe the energy intensity of the component at each moment, the instantaneous phase is used to characterize the phase travel of the component, and the instantaneous angular frequency is used to represent the instantaneous rate of change of the component.

[0008] In one embodiment, the step of extracting instantaneous features for each of the intrinsic mode function components includes: Short-term predictions are constructed for each of the intrinsic mode function components; The short-term predictions are synthesized to form the final short-term power prediction.

[0009] In one embodiment, the method further includes: Real-time calculation of photovoltaic power change rate; When the power change rate exceeds a preset first threshold, the value of the fluctuation suppression weight is increased; When the power change rate is lower than the preset first threshold, the value of the battery life optimization weight is increased.

[0010] In one embodiment, the energy storage system includes a supercapacitor and a lithium battery, and the method further includes: Obtain the duration of photovoltaic power fluctuations; If the duration is less than or equal to a preset time, the photovoltaic power fluctuation is responded to by the supercapacitor. If the duration is greater than the preset time, the photovoltaic power fluctuation is responded to by the lithium battery.

[0011] In one embodiment, controlling the energy storage system to perform coordinated charging and discharging includes: A distributed control strategy based on a consensus algorithm is adopted to control the state variables of each energy storage unit in the energy storage system.

[0012] In one embodiment, the method further includes: A simplified multi-objective function model is constructed based on the photovoltaic power prediction value. The simplified multi-objective function model is used to optimize power smoothing and battery life. Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, the simplified multi-objective function model is solved to obtain the power allocation command.

[0013] Secondly, embodiments of this application provide a control system for an energy storage system used in photovoltaic modules, applied to execute the energy storage system control method for photovoltaic modules as described in the first aspect, the system comprising: The data acquisition module is used to collect the real-time output power signal and related operating environment data of the photovoltaic module; The data processing module is used to perform empirical mode decomposition on the output power signal to obtain several intrinsic mode function components and residuals; The model building module is used to construct a multi-objective function model based on the photovoltaic power prediction value. The multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life and system operating cost of photovoltaic modules. The system control module is used to solve the multi-objective optimization model under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints to obtain power allocation instructions; and to control the energy storage system to perform coordinated charging and discharging according to the power allocation instructions.

[0014] Thirdly, embodiments of this application provide an electronic device, including: processor; Memory used to store processor-executable instructions; The processor is configured to implement the energy storage system control method for photovoltaic modules as described in the first aspect when executing the instructions.

[0015] This application provides a control method, system, and electronic equipment for an energy storage system used in photovoltaic modules. It achieves accurate short-term prediction of photovoltaic power through empirical mode decomposition and Hilbert transform, providing a forward-looking basis for subsequent control. Based on this prediction, a multi-objective optimization model is constructed to simultaneously coordinate three objectives: power fluctuation suppression, battery life extension, and system operating cost. Under the conditions of satisfying real-time power balance and energy storage state constraints, it generates optimal power allocation commands. This method can significantly improve the smoothness of photovoltaic grid-connected power, effectively reducing the impact on the grid; it delays battery capacity decay and extends the lifespan of the energy storage system by optimizing charging and discharging strategies; and it also considers economical operation, reducing the overall levelized cost of electricity (LCOE), achieving synergistic optimization of photovoltaic energy storage systems in terms of high efficiency, long lifespan, and low cost. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of an energy storage system control method for photovoltaic modules provided in an embodiment of this application.

[0017] Figure 2 A schematic diagram of a control system module for an energy storage system used in a photovoltaic module is provided as an embodiment of this application.

[0018] Figure 3 A schematic diagram of an electronic device provided in an embodiment of this application.

[0019] Explanation of main component symbols Control system for energy storage system used in photovoltaic modules 10 Data acquisition module 11 Data processing module 12 Model building module 13 System control module 14 Electronic devices 20 Processor 21 Memory 22 Method steps S100-700 Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0021] It should be noted that, in the embodiments of this application, "at least one" refers to one or more, and "more than one" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the specification of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application.

[0022] It should be noted that in the embodiments of this application, the terms "first," "second," etc., are used only for descriptive purposes and should not be construed as indicating or implying relative importance, nor as indicating or implying order. Features specified as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0023] Based on the embodiments described in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] As the global energy structure accelerates its transition towards cleaner and lower-carbon energy, photovoltaic (PV) power generation has become a significant growth driver due to its fuel-free and renewable nature. However, as the proportion of PV installations increases, its inherent intermittency and rapid time-varying characteristics pose new challenges to grid stability and system economics. Changes in sunlight (such as rapid cloud cover) can cause a sudden drop in array output (up to 50% or more) within seconds, while the grid's frequency and voltage regulation requirements often require millisecond-level responses. Traditional centralized energy storage and control systems often struggle to track and suppress such second-level disturbances due to communication delays (typical SCADA delays of approximately 50–100 ms) and the limited ramp-up rates of energy storage devices, leading to risks such as frequency deviations, harmonics, or protection malfunctions. Simultaneously, frequent charging and discharging accelerates lithium battery capacity degradation (cycle life decreases significantly at high depths of discharge), driving up PV costs and impairing long-term economic viability.

[0025] In related technologies, existing solutions mainly focus on three routes: First, achieving smooth output and grid connection through centralized or node-level energy storage in conjunction with inverters / PCS (but there are problems such as long response paths and cumulative energy conversion losses at multiple levels, with the overall system efficiency typically in the range of about 88% to 92%); Second, improving generation tracking and short-term output prediction capabilities by adopting improved MPPT and predictive control methods, but traditional MPPT and prediction methods still have limitations under multi-peak curves, local shading, and non-stationary signals; Third, attempting to balance transient response and energy support through hybrid energy storage (such as supercapacitors + lithium batteries) and distributed / module-level collaborative control, but lacking a unified multi-objective optimization framework in terms of capacity allocation, power scheduling strategies, lifetime optimization, and real-time trade-offs (fluctuation suppression vs. battery life vs. operating costs), and existing communication and coordination protocols, while ensuring fast response and reliability, often face problems of limited computing resources and insufficient solution speed. Based on the above shortcomings, there is an urgent need for a systematic solution that integrates physical design, second-level prediction, and hierarchical hybrid energy storage coordinated control to achieve safe, economical, and efficient grid connection of high-penetration photovoltaic systems.

[0026] This application provides a method, system, and electronic device for controlling an energy storage system for photovoltaic modules. By deeply coupling the energy storage unit with the photovoltaic module and realizing localized control at the module level, the transmission path of energy flow and control signals is significantly shortened, achieving millisecond-level response capability. This allows for rapid adjustment of output power in the event of sudden changes in illumination or partial shading, effectively suppressing frequency and voltage fluctuations on the grid-connected side, and improving the dynamic stability and anti-disturbance capability of the entire photovoltaic array during grid-connected operation.

[0027] Figure 1 This is a schematic flowchart of an energy storage system control method for photovoltaic modules provided in an embodiment of this application, as shown below. Figure 1 The control method for an energy storage system for photovoltaic modules shown includes at least the following steps: S100: Acquiring real-time output power signals and related operating environment data of the photovoltaic modules; S200: Performing empirical mode decomposition on the output power signals to obtain several intrinsic mode function components and residuals; S300: Performing Hilbert transform on each intrinsic mode function component to extract the instantaneous features of each intrinsic mode function component; S400: Synthesizing the predicted photovoltaic power value for future times based on the trend prediction of the instantaneous features and residuals of the intrinsic mode function components; S500: Constructing a multi-objective function model based on the predicted photovoltaic power value; S600: Solving the multi-objective optimization model under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints to obtain power allocation commands; S700: Controlling the energy storage system to perform coordinated charging and discharging according to the power allocation commands.

[0028] S100: Collects real-time output power signals and related operating environment data of photovoltaic modules.

[0029] In this embodiment of the application, the energy storage system control method for photovoltaic modules includes collecting the real-time output power signal of the photovoltaic modules and related operating environment data in step S100.

[0030] Specifically, in this embodiment, voltage and current sensors, as well as irradiance and temperature sensors, are arranged at the component level, with a sampling frequency of... (Example: 1kHz) Synchronously collect the voltage, current, and instantaneous power of the photovoltaic module. The system collects signals such as component temperature and irradiance; it first performs detrending and bandpass / lowpass filtering (to remove power frequency interference and high frequency noise) on the raw collected signals, and caches them in a sliding window manner for subsequent EMD decomposition and short-term prediction; the controller timestamps the sampled data and saves several historical windows for model training and online self-calibration.

[0031] Understandably, the sampling rate, filter parameters, and window length mentioned above can be optimized and adjusted according to specific hardware performance and on-site operating conditions; "Relevant operating environment data" may also include inverter operating status, local load information, and the status of neighboring components, in order to be used for more accurate prediction and scheduling.

[0032] S200: Perform empirical mode decomposition on the output power signal to obtain several intrinsic mode function components and residuals.

[0033] In the embodiments of this application, the energy storage system control method for photovoltaic modules includes performing empirical mode decomposition on the output power signal in step S200 to obtain several intrinsic mode function components and residuals.

[0034] Specifically, the filtered instantaneous power signal is processed within each sliding window. Apply Empirical Mode Decomposition (EMD) to generate IMF component sets. With residual ,satisfy:

[0035] in, For output power signal, Indicates the first Each intrinsic mode function component This is the residual term. The termination criterion for EMD can be the difference threshold of the number of extreme points, the residual signal energy threshold, or the maximum number of iterations. In order to ensure real-time performance, EMD can be truncated within a finite number of extreme points or a fixed time budget and output an approximate IMF.

[0036] Understandably, the number of IMFs The accuracy of the decomposition is affected by the window length, noise level, and termination criteria. During implementation, a suitable field-appropriate decomposition method can be determined offline using historical data. The scope is designed to balance accuracy and computational cost.

[0037] S300: Perform Hilbert transform on each intrinsic mode function component to extract the instantaneous features of each intrinsic mode function component.

[0038] In the embodiments of this application, the energy storage system control method for photovoltaic modules includes performing a Hilbert transform on each intrinsic mode function component in step S300 to extract the instantaneous features of each intrinsic mode function component.

[0039] Specifically, a Hilbert transform is performed on each intrinsic mode function (IMF) component to extract its instantaneous features. For each IMF... Applying Hilbert transform (This can be achieved using principal value integration or numerical Hilbert filtering), and the instantaneous amplitude and instantaneous phase can be calculated accordingly:

[0040] This leads to the instantaneous frequency. The instantaneous amplitude With instantaneous frequency Used to characterize the energy and rate of change of each IMF at that moment, and as input features for short-term prediction models.

[0041] Understandably, the Hilbert transform described above can be implemented in the discrete Fourier domain to reduce computational complexity, and smoothing or regularization can be applied to noise-sensitive IMFs to improve the robustness of instantaneous features.

[0042] S400: Based on the instantaneous characteristics of the intrinsic mode function components and the trend prediction of the residuals, the predicted value of photovoltaic power at future times is synthesized.

[0043] In the embodiments of this application, the energy storage system control method for photovoltaic modules includes, in step S400, synthesizing a predicted value of photovoltaic power at a future time based on the instantaneous characteristics and residual trend prediction of the intrinsic mode function components.

[0044] Specifically, short-term prediction models (such as low-order autoregressive models, lightweight LSTM, or small fully connected networks) are established for several selected IMF components to predict... For the residual term Obtained by trend extrapolation or linear / quadratic fitting The confidence levels for short-term power prediction and output prediction are synthesized according to the following formula. : Prediction confidence It can be obtained from historical residual variance, model uncertainty estimation, or cross-validation error estimation, and used as the basis for subsequent optimization weight adjustment.

[0045] Understandably, the prediction window The selection of the time (e.g., 0.1s–10s) should be combined with the target control delay and energy storage response capability; in low confidence situations, the system can amplify the lifetime weight or trigger a more conservative scheduling strategy.

[0046] S500: Construct a multi-objective function model based on the photovoltaic power prediction values.

[0047] In this embodiment of the application, the energy storage system control method for photovoltaic modules includes, in step S500, constructing a multi-objective function model based on the photovoltaic power prediction value.

[0048] Specifically, the multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life, and system operating cost of photovoltaic modules. The expression of the multi-objective function is as follows:

[0049] in, For fluctuation suppression weights, This represents the actual power fluctuation value. Rated power of photovoltaic modules Optimize weights for battery life. This represents the actual number of loops. This is the maximum design life of the energy storage device. As a weight for system operating costs, This represents an optimized running cycle. Represented as a time period; Indicates the economic coefficient of photovoltaic power generation; This indicates the maintenance cost of the energy storage system; This indicates that the system trades subsidies to the power grid; This refers to subsidies provided in response to load shifting.

[0050] It is understandable that the economic coefficients in the above objective function can be assigned values ​​according to station-level or field-level strategies, and can be dynamically adjusted at different times (such as peak and off-peak electricity prices) to balance short-term stability and long-term returns.

[0051] S600: Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, solve the multi-objective optimization model to obtain the power allocation command.

[0052] In the embodiments of this application, the energy storage system control method for photovoltaic modules includes, in step S600, solving a multi-objective optimization model to obtain a power allocation command under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints.

[0053] Specifically, the optimization solution needs to satisfy real-time power balance constraints and SOC update constraints, for example:

[0054]

[0055] And satisfy the boundary conditions , , To meet real-time requirements, the solver can employ a lightweight heuristic algorithm (such as PSO) and limit the maximum number of iterations and computation time budget; the solution result is deterministic. and Assignment command (or equivalent current command).

[0056] It is understandable that constraint penalties or repair strategies can be used during the solution process to ensure that the obtained solution meets the safety boundary, and station / field level instructions can be introduced as soft constraints when necessary to achieve multi-level collaboration.

[0057] S700: Controls the energy storage system to perform coordinated charging and discharging according to power distribution commands.

[0058] In this embodiment of the application, the energy storage system control method for photovoltaic modules includes controlling the energy storage system to perform coordinated charging and discharging according to the power distribution command in step S700.

[0059] Specifically, the component-level controller will allocate and The command is converted into a power or current command for the bidirectional DC / DC converter and inverter / PCS and then issued for execution. Before the command is issued, the temperature, voltage, SOC and device ratings must be checked by the safety verification module, and the command rate is limited to prevent sudden step changes from causing device damage. During the execution process, the execution feedback is collected in real time and the log is recorded, the SOC and energy metering are updated, and abnormal or alarm information is reported to the field / station level or the cloud.

[0060] Understandably, when communication fails, central control loses connection, or device malfunction is detected, the component-level controller automatically switches to local degraded control mode: based on local prediction... With enhanced lifetime weights as input, the local optimization problem is solved again, the discharge depth is limited, and the supercapacitor is given priority to handle the high-frequency transient response. At the same time, the fault diagnosis process is triggered to isolate suspected faulty units, and the neighboring components take over the short-term tasks until maintenance is carried out.

[0061] The embodiments of this application adopt a hierarchical collaborative control and transient / steady-state energy task allocation strategy, which can organically combine the characteristics and advantages of supercapacitors and lithium batteries: the supercapacitors bear the high-frequency short-time power surges, and the lithium batteries are responsible for energy support and energy balance, reducing the deep charge and discharge of the battery under high-frequency cycling, thereby reducing the capacity decay rate, extending the overall service life of the energy storage unit and improving the long-term reliability of the system.

[0062] Thanks to module-level energy storage and intelligent control strategies, each photovoltaic module can independently perform power regulation and local energy balancing. This not only reduces power loss caused by local shading but also improves the overall utilization rate and grid connection quality of photovoltaic power generation. The system can achieve local optimization without relying on a central node for all decisions, reducing dependence on centralized communication links and maintaining basic power regulation capabilities in the event of network anomalies or communication interruptions.

[0063] This application's embodiments design a lightweight scheduling and prediction module for real-time and resource-constrained scenarios, balancing control accuracy and computational complexity. This makes the solution easy to deploy on edge processors or component-level controllers, with good scalability and engineering feasibility, facilitating its widespread application in distributed large-scale photovoltaic scenarios.

[0064] In summary, this application has significant advantages in improving grid-connected dynamic performance, extending energy storage life, increasing power generation utilization, and reducing operation and maintenance and life cycle costs, and can provide strong support for the safe, economical, and stable grid connection of high-penetration photovoltaic systems.

[0065] In this embodiment, empirical mode decomposition (EMD) of the output power signal includes: identifying all local extrema of the output power signal; constructing an upper and lower envelope using cubic spline interpolation based on the local extrema; calculating the envelope mean of the upper and lower envelopes; subtracting the envelope mean from the output power signal to obtain the first undetermined intrinsic mode function (EMF) component; determining whether the first undetermined EMF component satisfies the conditions for an EMF; if so, recording the first undetermined EMF component as the first EMF component; separating the first EMF component from the output power signal to obtain the first residual term; iterating the residual term as a new output power signal until the final residual term is a monotonic function or a constant, thus obtaining at least one EMF component and the residual term.

[0066] Specifically, the identification of local extrema can be achieved by calculating the first-order difference of the power signal within a sliding window and detecting sign changes. Mirror extension or additional extrema points are used at the endpoints to avoid boundary oscillations in cubic spline interpolation. Both the upper and lower envelopes use cubic spline interpolation, and the convergence of the envelope mean is checked in each iteration (e.g., the change in the envelope mean is considered convergent if it is below a preset threshold). When determining whether a component satisfies the conditions for an intrinsic mode function (IMF), a common criterion is used—the number of extrema and the number of zero crossovers of the component differ by no more than 1, and its local envelope mean is approximately zero (the absolute value of the mean is less than a threshold). If these conditions are not met, the component continues to undergo sifting until it meets the criteria or the maximum number of sifting iterations is reached. The residual iteration terminates when the residual is a monotonic function or an approximate constant, or when the energy of the residual is lower than a small proportion of the total energy.

[0067] It is understandable that the above-mentioned threshold, maximum number of filtering times and window length can be adjusted according to the on-site signal noise level and control real-time requirements; on resource-constrained controllers, approximate EMD or pre-calculated templates can be used to reduce the computational burden.

[0068] In this embodiment, a Hilbert transform is performed on each intrinsic mode function (EMF) component to obtain the imaginary part of the corresponding analytic signal. Based on this analytic signal, the instantaneous amplitude, instantaneous phase, and instantaneous angular frequency corresponding to each EMF component are constructed. The instantaneous amplitude describes the energy intensity of the component at each moment, the instantaneous phase characterizes the phase travel time of the component, and the instantaneous angular frequency represents the instantaneous rate of change of the component. The expression for the Hilbert transform of each EMF component is as follows:

[0069]

[0070]

[0071]

[0072] in, For the Hilbert transform operator, Indicates the first Individual eigenmode functions The result obtained after performing the Hilbert transform constitutes the imaginary part of the analytic signal. Indicates the first The instantaneous amplitude of each eigenmode function. Indicates the first The instantaneous phase of each eigenmode function, Indicates the first The instantaneous angular frequency of each eigenmode function. Represents Cauchy's principal value integral. For integration variables, It is a time variable.

[0073] Specifically, in practice, the frequency domain method of FFT / inverse FFT or numerical Hilbert filter can be used to approximate the Hilbert transform, and the phase derivative can be smoothed by difference or polynomial fitting to suppress numerical noise. For short pulses or low-energy IMFs, amplitude threshold filtering should be performed first to avoid misleading the instantaneous frequency calculation.

[0074] Understandably, the numerical implementation of the Hilbert transform needs to pay attention to the endpoint effect and filter group delay. In engineering implementation, a mirror expansion can be performed before and after the IMF, and short-term smoothing can be applied to the obtained instantaneous features to improve robustness.

[0075] In this embodiment, the instantaneous features of each intrinsic mode function component are extracted, including: constructing short-term predictions for each intrinsic mode function component and synthesizing a final short-term power prediction.

[0076] in, For the final short-term power prediction, For the prediction window, For residual prediction, For short-term forecasting, forecast confidence level It is obtained from the residual variance or model uncertainty estimation.

[0077] Specifically, low-order autoregressive models (AR(p)) can be used for high-frequency IMFs, while lightweight LSTM or small feedforward neural networks can be used for medium- and low-frequency IMFs. Residual terms can be estimated using linear extrapolation or polynomial fitting. Model training and hyperparameter selection are completed through offline cross-validation using historical data, and sliding window incremental updates are used during online operation to adapt to component performance degradation. Prediction confidence. Can be defined as Or it can be calculated based on Bayesian confidence intervals, where This represents the residual variance.

[0078] Understandably, the specific selection and complexity of the prediction model should balance edge computing capabilities and real-time requirements. In situations where resources are limited, low-rank models can be prioritized and predictions can be made only for a few IMFs that can significantly contribute to the prediction accuracy.

[0079] In this embodiment of the application, the method further includes: calculating the photovoltaic power change rate in real time. When the power change rate exceeds a preset first threshold, the value of the fluctuation suppression weight is increased. When the power change rate is lower than the preset first threshold, the value of the battery life optimization weight is increased.

[0080] Specifically, the real-time power change rate can be calculated as follows: ,in Set a threshold for a short time difference window (e.g., 0.1s). ,when And prediction confidence When the weight is below the threshold, (Suppression of fluctuations) Proportional amplification (e.g.) ); on the contrary when And when the system is in steady state, the improvement Prioritize reducing deep discharge. Parameters , It can be determined through engineering experience and simulation.

[0081] Understandably, the threshold and amplification ratio can be adjusted according to the season, component degradation, or grid connection rules; in extreme short-pulse scenarios, the weight can be temporarily switched to a strategy where the supercapacitor bears the entire weight to quickly suppress fluctuations.

[0082] In this embodiment, the energy storage system includes a supercapacitor and a lithium battery. The method further includes: obtaining the duration of photovoltaic power fluctuations. If the duration is less than or equal to a preset time, the photovoltaic power fluctuations are responded to by the supercapacitor. If the duration is greater than the preset time, the photovoltaic power fluctuations are responded to by the lithium battery.

[0083] Specifically, define the pulse duration threshold. (For example, 100ms, determined based on the characteristics of the supercapacitor and battery), measuring the duration of the event when a single fluctuation event is detected. :like Then, in power allocation, transient components will be... Primarily assigned to supercapacitors (i.e., to maximize the use of supercapacitors) (bearing high-frequency energy); if Then the steady-state energy compensation will be allocated to the battery (increasing) The system can also set cumulative event counts and cooldown periods to avoid oscillations caused by frequent switching.

[0084] Understandably, the threshold The allocation strategy can be determined based on the supercapacitor's energy capacity, equivalent series resistance, and the battery's allowable discharge rate, and can be dynamically adjusted according to the health status during operation.

[0085] In this embodiment, controlling the energy storage system to perform coordinated charging and discharging includes: employing a distributed control strategy based on a consensus algorithm to control the state variables of each energy storage unit in the energy storage system. The expression for the consensus algorithm is:

[0086] in, For the first The energy storage unit in the first The state variables of each control cycle For the first Component state variables For elements of the adjacency matrix, For the first The set of neighbors of each component To converge the step size.

[0087] Specifically, state variables This can be SOC, allocation coefficient, or local power error; adjacency weight matrix Designed as a symmetric or column random matrix to ensure consistent convergence; convergence step size The value must satisfy the spectral radius constraint to ensure stability (e.g.) ,in (This refers to the Laplace matrix of the graph). In the event of packet loss or delay in the link, asynchronous consistency or event-triggered mechanisms are used to improve robustness.

[0088] Understandably, the distributed consensus mechanism enables components to reach coordinated decisions without the need for large-scale computation by a central node, which helps avoid single points of failure and reduce communication bandwidth requirements; global policies can still be periodically issued at the field level or in the cloud to correct local deviations.

[0089] In this embodiment, the method further includes: constructing a simplified multi-objective function model based on the photovoltaic power prediction value; the simplified multi-objective function model is used to optimize power smoothing and battery life. Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, the simplified multi-objective function model is solved to derive the power allocation command. The expression of the simplified multi-objective function is:

[0090] in, Weights for fluctuation suppression; Optimize weights for lifespan; The rate of change of power; No. Deep discharge in one charge-discharge cycle.

[0091] Specifically, in the discretization implementation, the integral term is approximated as... Alternatively, it can be in the form of a sum of squares to facilitate quadratic programming solutions; DOD quantization can be based on the ratio of energy in a single cycle to the rated energy and recorded in a cycle counter; fast convex optimization or constrained heuristic solutions can be used on the edge controller to return executable allocation instructions in each control cycle.

[0092] Understandably, simplified objectives are suitable for scenarios with high real-time requirements and limited edge resources. If there are more computing resources at the station level or in the cloud, a full multi-objective model combined with economic terms can be used for long-term optimization.

[0093] The following describes an exemplary embodiment of the energy storage system control method for photovoltaic modules provided in this application.

[0094] This embodiment uses a rooftop distributed photovoltaic (PV) system in an industrial park as an application scenario. The installed capacity is approximately 100kW, using 182 PV modules with a rated power of 550W (100000÷550≈181.82, rounded to 182) connected to the grid. Each PV module adopts an integrated design, encapsulating the PV cells, hybrid energy storage unit (supercapacitor + LFP battery), bidirectional DC / DC converter, module-level inverter / PCS, and module-level intelligent controller in the same back box, achieving the shortest physical energy flow and control signal path. Exemplary hardware specifications are: PV module rated power 550W; supercapacitor nominal 48V / 165F (for millisecond-level transient support); LFP battery branches are combined in series and parallel at 3.2V / 50Ah to match the system voltage (for steady-state energy support); the controller example is an ARM Cortex-M7 (400MHz) + FPGA coprocessor with a sampling rate up to 1kHz; communication uses power line carrier or lightweight wireless links to ensure reliable switching under low bandwidth. For engineering reliability considerations, the package includes a thermally conductive silicone pad, a top heat sink, and an IP65 protection rating, and provides temperature, overcurrent, and overvoltage protection at the electrical level.

[0095] In terms of algorithm implementation, the component-level controller first synchronously collects operating and environmental data such as voltage, current, component temperature, and irradiance at a high sampling rate and then calculates the instantaneous power sequence. To separate the fluctuation components at different time scales, the following methods were employed: Perform Empirical Mode Decomposition (EMD):

[0096] in For the first One intrinsic mode function (IMF). This represents the residual term (low-frequency trend or noise). EMD is implemented by identifying local extrema, constructing upper / lower envelopes using cubic spline interpolation, and performing a sifting operation to calculate the envelope mean to obtain the IMF, stopping when the residual becomes a monotonic function or the energy falls below a threshold. A Hilbert transform is further applied to each IMF to extract instantaneous features.

[0097]

[0098] in Characterizes instantaneous amplitude (energy intensity). It represents the instantaneous angular frequency (rate of change). To ensure numerical stability, mirror spread and frequency domain Hilbert filtering can be used in the implementation, and the phase derivative can be smoothed.

[0099] Based on the instantaneous amplitude and frequency characteristics of each IMF and the trend extrapolation of the residuals, a short-term prediction model is constructed and the final short-term power prediction is synthesized:

[0100] in For the prediction window (example: 0.1s–10s), and These are short-term predictions for IMF and residuals, respectively. For high-frequency IMFs, a low-order autoregressive (AR) model can be used; for low-frequency IMFs or residuals, a lightweight LSTM or small neural network can be used. The prediction confidence level is... The results are derived from the residual variance or Bayesian uncertainty, serving as the basis for subsequent weight adaptation.

[0101] A multi-objective optimization model is constructed using the prediction results to dynamically balance fluctuation suppression, battery life, and operating costs. The objective function is expressed as follows:

[0102] The meanings of each symbol are as follows: This represents the actual power fluctuation during the forecast period. Rated power of the component; To accumulate the number of loops, The number of cycles corresponding to the design life; These are the power generation economic coefficient, energy storage operation and maintenance cost, grid transaction / subsidy, and load response subsidy items; weights Based on the prediction confidence level With real-time power change rate Adaptive adjustment, such as amplifying fluctuation weights according to a relational formula:

[0103] in As the initial value, For adjustment coefficients, This is the power change rate threshold.

[0104] The optimization solution needs to satisfy real-time power balance and SOC constraints. The discrete updates of power balance and SOC are as follows:

[0105] and satisfy , , To meet the real-time requirements at the edge, the solver employs lightweight heuristic algorithms (such as particle swarm optimization, PSO) or fast convex approximation, and limits the maximum number of iterations and time budget.

[0106] To quantify the design of supercapacitors and batteries, the capacitance required for a supercapacitor must satisfy the energy demand formula:

[0107] Therefore, we can conclude that: In the formula To design pulse power, For duration, For loop efficiency, This refers to the operating voltage. The battery's rated capacity is determined based on steady-state energy compensation requirements.

[0108] in For steady-state energy demand (Wh). and These are the nominal voltage and efficiency, respectively; when selecting a model, also refer to the battery's DOD-life curve to determine the required... .

[0109] In terms of distributed / field-level coordination, each component can use a consensus algorithm to coordinate state and allocation coefficients: Let the components... State variables are Then the consistent update is

[0110] The adjacency matrix With step size The design ensures convergence; this distributed strategy can make local decisions more consistent without relying on a central node, which is beneficial for fault tolerance and scalability.

[0111] The implementation also specifies an event recognition and degradation strategy: the controller calculates the power change rate in real time. and duration of a single event ,like and (Supercapacitor time-domain threshold), then the supercapacitor response preferentially suppresses high-frequency pulses; if In this case, the battery will bear the steady-state energy. When communication is lost or a device malfunctions, the component-level controller automatically switches to a local conservative strategy to increase lifetime weight and limit the depth of discharge. The fault diagnosis module determines the anomaly and isolates the faulty unit by predicting the actual residual, single-cell voltage mutation and internal resistance rise. The task is temporarily taken over by a neighboring component or supercapacitor until maintenance is carried out.

[0112] With similar configurations, supercapacitor response can reach approximately 5–10 ms, peak transient suppression efficiency can reach 70%–85%, the equivalent cycle load of hybrid energy storage is significantly reduced, thereby increasing battery life by approximately 30%–50%, and the system-level energy conversion efficiency is close to 94%–96% (depending on the specific device and converter efficiency). The above figures are exemplary engineering results; actual performance should be precisely calculated based on on-site irradiation conditions, inter-module shading modes, device characteristics, and parameter settings.

[0113] In summary, this exemplary embodiment combines integrated hardware design (component-level hybrid energy storage + high-performance edge controller + bidirectional converter) with short-term prediction, multi-objective optimization and distributed consensus collaborative control based on EMD-Hilbert, achieving a comprehensive improvement in millisecond / second-level suppression of grid-connected power, battery life protection and operational economy under highly variable illumination conditions.

[0114] Figure 2 This is a schematic diagram of a control system module for an energy storage system used in photovoltaic modules, provided in one embodiment of this application. Figure 2 The energy storage system control system 10 shown for photovoltaic modules includes at least the following modules: data acquisition module 11, data processing module 12, model building module 13, and system control module 14.

[0115] In this embodiment, the data acquisition module 11 is used to acquire the real-time output power signal of the photovoltaic module and related operating environment data. Please refer to the documentation for specific acquisition methods. Figure 1 The details and their corresponding descriptions are not repeated here.

[0116] In this embodiment, the data processing module 12 is used to perform empirical mode decomposition on the output power signal to obtain several intrinsic mode function components and residuals. The formula for empirical mode decomposition is as follows:

[0117] in, For output power signal, Indicates the first Each intrinsic mode function component The residual term is used; a Hilbert transform is performed on each intrinsic mode function (IMF) component to extract its instantaneous features, including instantaneous frequency and instantaneous amplitude; based on the instantaneous features of the IMF components and the trend prediction of the residual, the predicted photovoltaic power value for future times is synthesized. Please refer to the following for specific acquisition methods. Figure 1 The details and their corresponding descriptions are not repeated here.

[0118] In this embodiment, the model building module 13 is used to construct a multi-objective function model based on the photovoltaic power prediction value. The multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life, and system operating cost of the photovoltaic module. The expression of the multi-objective function is:

[0119] in, For fluctuation suppression weights, This represents the actual power fluctuation value. Rated power of photovoltaic modules Optimize weights for battery life. This represents the actual number of loops. This is the maximum design life of the energy storage device. As a weight for system operating costs, This represents an optimized running cycle. Represented as a time period; Indicates the economic coefficient of photovoltaic power generation; This indicates the maintenance cost of the energy storage system; This indicates that the system trades subsidies to the power grid; This refers to subsidies provided in response to load shifting. Please refer to the attached document for details on how to obtain these subsidies. Figure 1 The details and their corresponding descriptions are not repeated here.

[0120] In this embodiment, the system control module 14 is used to solve a multi-objective optimization model under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints to obtain power allocation commands; and to control the energy storage system to perform coordinated charging and discharging according to the power allocation commands. Please refer to the following for specific control methods. Figure 1 The details and their corresponding descriptions are not repeated here.

[0121] Figure 3 This is an electronic device 20 provided in one embodiment of this application. For example... Figure 3 As shown, the electronic device 20 includes at least the following components: a processor 21 and a memory 22.

[0122] In this embodiment, the memory 22 is used to store executable instructions of the processor 21, which, when configured to execute instructions, implement... Figure 1The method for controlling an energy storage system for photovoltaic modules is shown.

[0123] In one embodiment of this application, the program operating in the electronic device 20 may be a program that controls a central processing unit (CPU) or similar device to achieve the functions of the above-described embodiments of the present invention (a program that enables the computer to function). The information processed by these devices is then temporarily stored in random access memory (RAM) during processing, and subsequently stored in various ROMs such as read-only memory (Flash ROM) and hard disk drives (HDDs), and read, corrected, and written by the CPU as needed.

[0124] It should be noted that a portion of the electronic device 20 described above can also be implemented using a computer. In this case, the program for implementing the control function can be recorded on a computer-readable recording medium, and the program recorded on the recording medium can be read into the computer and executed.

[0125] It should be noted that the term "computer" as used here refers to a computer built into electronic device 20, employing hardware including an operating system and peripheral devices. Furthermore, "computer-readable recording media" refers to removable media such as floppy disks, magneto-optical disks, ROMs, and CD-ROMs, as well as storage devices such as hard drives built into the computer.

[0126] Furthermore, a "computer-readable recording medium" can include: a medium that dynamically stores a program for a short period of time, such as a communication line used when transmitting a program via a network such as the Internet or a communication line such as a telephone line; or a medium that stores a program for a fixed period of time, such as volatile memory inside a computer that serves as a server or client in this case. In addition, the aforementioned program can be a program used to implement the above-mentioned functions, or it can be a program that can implement the above-mentioned functions by combining with programs already recorded in the computer.

[0127] Furthermore, the electronic device 20 in the above embodiments can also be implemented as an assembly (device group) composed of multiple devices. Each device constituting the device group can possess some or all of the functions or functional blocks of the electronic device 20 in the above embodiments. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device 20.

[0128] It is understood that the energy storage system control method, system, and electronic equipment for photovoltaic modules provided in this application achieve accurate short-term prediction of photovoltaic power through empirical mode decomposition and Hilbert transform, providing a forward-looking basis for subsequent control. Based on this prediction, a multi-objective optimization model is constructed to simultaneously coordinate three objectives: power fluctuation suppression, battery life extension, and system operating cost. Under the condition of satisfying real-time power balance and energy storage state constraints, it generates the optimal power allocation command. This method can significantly improve the smoothness of photovoltaic grid-connected power and effectively reduce the impact on the grid; it delays battery capacity decay and extends the lifespan of the energy storage system by optimizing charging and discharging strategies; and it also takes into account economic operation, reducing the overall levelized cost of electricity, thus achieving synergistic optimization of photovoltaic energy storage systems in terms of high efficiency, long lifespan, and low cost.

[0129] Those skilled in the art should recognize that the above embodiments are only used to illustrate this application and are not intended to limit this application. Any appropriate changes and variations made to the above embodiments within the essential spirit and scope of this application fall within the scope of protection claimed in this application.

Claims

1. A control method for an energy storage system in photovoltaic modules, applied to controlling photovoltaic modules with an energy storage system, characterized in that, The method includes: Collect the real-time output power signal and related operating environment data of the photovoltaic module; Empirical mode decomposition is performed on the output power signal to obtain several intrinsic mode function components and residuals; A Hilbert transform is performed on each of the intrinsic mode function components to extract the instantaneous features of each intrinsic mode function component, the instantaneous features including instantaneous frequency and instantaneous amplitude; Based on the instantaneous characteristics of the intrinsic mode function components and the trend prediction of the residuals, the photovoltaic power prediction value for future moments is synthesized, and the prediction confidence level is obtained. A multi-objective function model is constructed based on the photovoltaic power prediction value. The multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life and system operating cost of photovoltaic modules. The photovoltaic power change rate is calculated in real time, and the weights in the multi-objective function model are adaptively adjusted based on the prediction confidence level and the photovoltaic power change rate; the adaptive adjustment of the weights includes: When the absolute value of the photovoltaic power change rate is greater than the preset power change rate threshold, the fluctuation suppression weight is amplified according to the following formula: ； in, Weights for fluctuation suppression As the initial value, For adjustment coefficients, To predict confidence levels, For real-time power change rate, The power change rate threshold; When the absolute value of the photovoltaic power change rate is less than or equal to the power change rate threshold and the system is in steady state operation, the value of the battery life optimization weight is increased. Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, the multi-objective optimization model is solved to obtain power allocation instructions; According to the power allocation command, the energy storage system is controlled to perform coordinated charging and discharging.

2. The energy storage system control method for photovoltaic modules according to claim 1, characterized in that, The empirical mode decomposition of the output power signal includes: Identify all local extrema of the output power signal; Using the aforementioned local extreme points, the upper and lower envelopes are constructed using cubic spline interpolation. Calculate the envelope mean of the upper envelope and the lower envelope; Subtract the envelope mean from the output power signal to obtain the first undetermined intrinsic mode function component; Determine whether the first undetermined intrinsic mode function component satisfies the conditions for intrinsic mode functions; If satisfied, the first undetermined intrinsic mode function component is recorded as the first intrinsic mode function component; The first intrinsic mode function component is separated from the output power signal to obtain the first residual term; The residual term is iterated repeatedly to form a new output power signal until the final residual term is a monotonic function or a constant, thereby obtaining at least one intrinsic mode function component and the residual term.

3. The energy storage system control method for photovoltaic modules according to claim 2, characterized in that, The step of performing a Hilbert transform on each of the intrinsic mode function components includes: A Hilbert transform is performed on each of the intrinsic mode function components to obtain the imaginary part of the corresponding analytic signal; Based on the analytical signal, the instantaneous amplitude, instantaneous phase, and instantaneous angular frequency corresponding to each of the intrinsic mode function components are constructed; The instantaneous amplitude is used to describe the energy intensity of the component at each moment, the instantaneous phase is used to characterize the phase travel of the component, and the instantaneous angular frequency is used to represent the instantaneous rate of change of the component.

4. The energy storage system control method for photovoltaic modules according to claim 3, characterized in that, The step of extracting the instantaneous features of each of the intrinsic mode function components includes: Short-term predictions are constructed for each of the intrinsic mode function components; The short-term predictions are synthesized to form the final short-term power prediction.

5. A control method for an energy storage system for photovoltaic modules according to claim 1, wherein the energy storage system comprises a supercapacitor and a lithium battery, characterized in that, The method further includes: Obtain the duration of photovoltaic power fluctuations; If the duration is less than or equal to a preset time, the photovoltaic power fluctuation is responded to by the supercapacitor. If the duration is greater than the preset time, the photovoltaic power fluctuation is responded to by the lithium battery.

6. The energy storage system control method for photovoltaic modules according to claim 5, characterized in that, The control of the energy storage system to perform coordinated charging and discharging includes: A distributed control strategy based on a consensus algorithm is adopted to control the state variables of each energy storage unit in the energy storage system.

7. The energy storage system control method for photovoltaic modules according to claim 6, characterized in that, The method further includes: A simplified multi-objective function model is constructed based on the photovoltaic power prediction value. The simplified multi-objective function model is used to optimize power smoothing and battery life. Under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints, the simplified multi-objective function model is solved to obtain the power allocation command.

8. A control system for an energy storage system for photovoltaic modules, applied to perform the control method for an energy storage system for photovoltaic modules as described in any one of claims 1 to 7, characterized in that, The system includes: The data acquisition module is used to collect the real-time output power signal and related operating environment data of the photovoltaic module; The data processing module is used to perform empirical mode decomposition on the output power signal to obtain several intrinsic mode function components and residuals; The model building module is used to construct a multi-objective function model based on the photovoltaic power prediction value. The multi-objective function model is used to optimize the energy storage system fluctuation suppression, battery life and system operating cost of photovoltaic modules. The system control module is used to solve the multi-objective optimization model under the conditions of satisfying power balance constraints and energy storage unit state of charge constraints to obtain power allocation instructions; and to control the energy storage system to perform coordinated charging and discharging according to the power allocation instructions.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the energy storage system control method for photovoltaic modules according to any one of claims 1 to 7 when executing the instructions.

Images