A wide temperature environment self-organizing adaptation-based all-vanadium redox flow energy storage system

Abstract

The application discloses a kind of based on wide temperature environment self-organizing adaptation's all-vanadium redox flow energy storage system, including all-vanadium redox flow battery module, multi-physical field perception network and all-temperature range self-organizing adaptation controller.Multiple physical field perception network obtains the data such as each valence state concentration of electrolyte, total vanadium concentration and the temperature field inside the stack by multi-source sensor in real time, and all-temperature range self-organizing adaptation controller integrates electrolyte state self-adaptive module and stack thermal field self-balancing module, electrolyte state self-adaptive module is based on physical information neural network Construction prediction model, according to the prediction result, actively adjust electrolyte temperature and circulating pump flow;Stack thermal field self-balancing module is accurately adjusted electrolyte flow distribution by the electrically controlled valve independently arranged in each branch flow channel, based on PID control algorithm, and realizes two-stage thermal management in combination with external liquid cooling system.The application can be stably operated under the wide temperature environment of-30~50 ℃, effectively eliminate local hot spots inside the stack, prolong the service life of system.

Description

Technical Field

[0001] This invention belongs to the field of electrochemical energy storage technology, specifically relating to an all-vanadium redox flow battery energy storage system, and in particular a system capable of self-organizing, self-adapting, and self-optimizing operation over a wide temperature range. Background Technology

[0002] Vanadium redox flow batteries, as a novel energy storage technology characterized by high safety, long cycle life, and independently designable capacity and power, have shown broad application prospects in areas such as renewable energy grid integration, grid peak shaving and frequency regulation, and user-side energy storage. As the global energy structure accelerates its transition to clean and low-carbon energy, the operating environment faced by energy storage systems is becoming increasingly complex and variable. This is particularly true for energy storage power stations deployed in scenarios such as the northwestern deserts, the northeastern cold regions, and offshore wind power, where the operating temperature range often spans -30 to 50°C, posing a severe challenge to the environmental adaptability of energy storage systems.

[0003] In existing vanadium redox flow battery technologies, operating temperature control typically employs relatively simple methods. Most systems rely on external thermal management systems for passive temperature regulation, such as activating heaters or cooling fans when sensors detect electrolyte temperatures exceeding preset thresholds. This control method suffers from significant lag, often only initiating adjustments after the temperature has already deviated from the normal range, making precise temperature management and proactive prevention difficult. More critically, the electrolyte exhibits drastically different physicochemical properties at different temperatures: viscosity increases sharply and vanadium ion solubility decreases at low temperatures, while high temperatures pose a risk of thermal precipitation. Existing systems lack the ability to precisely sense and dynamically respond to these complex temperature effects. Meanwhile, as the core component of the vanadium redox flow battery, the uniformity of internal temperature distribution directly affects the system's operating efficiency and cycle life. In actual operation, due to the difficulty in achieving completely uniform electrolyte distribution in parallel channels, coupled with unavoidable manufacturing differences between individual cells, localized hot spots often appear within the stack. If these hot spots are not effectively controlled over a long period, they will accelerate the oxidation and corrosion of the carbon felt electrodes, shorten the lifespan of the ion-exchange membrane, and in severe cases, may even pose safety hazards. However, existing technologies have limited means of monitoring the internal temperature field of fuel cell stacks. Most of them only set up a small number of temperature measuring points at the inlet and outlet of the fuel cell stack, which cannot know the true temperature distribution inside the fuel cell stack, let alone accurately locate and adjust local hot spots. Summary of the Invention

[0004] This invention provides a vanadium redox flow energy storage system based on self-organizing adaptation in a wide temperature environment, in order to solve the problems of poor adaptability to extreme temperature environments and low operating efficiency of existing technologies.

[0005] Specifically, the technical solution provided by this invention is as follows: A vanadium redox flow storage system based on self-organizing adaptation in a wide temperature range includes: Vanadium redox flow battery module; Multiphysics sensing networks are used to acquire multi-dimensional state information of system operation in real time; A full-temperature-range self-organizing adaptive controller is connected to the vanadium redox flow battery module and the multi-physics sensing network, respectively. The full-temperature-range self-organizing adaptive controller includes an electrolyte state adaptive module and a stack thermal field self-balancing module. The electrolyte state adaptive module is used to receive electrolyte state information collected by the multi-physics sensing network, predict the future state evolution trend of the electrolyte based on the physical information neural network model, and generate temperature adjustment command and circulation pump flow rate adjustment command according to the prediction results, and send them to the vanadium redox flow battery module respectively. The self-balancing module for the thermal field of the fuel cell stack is used to receive the internal temperature field information of the fuel cell stack collected by the multi-physics sensing network, and adjust the electrolyte flow distribution of each branch channel according to the deviation between the measured temperature and the target temperature of each region. At the same time, it adjusts the working state of the external liquid cooling system according to the overall temperature condition of the fuel cell stack.

[0006] Furthermore, the vanadium redox flow battery module includes a stack unit, an electrolyte storage tank unit, a circulation pump group, a piping system, and a heat exchange unit. The stack unit is formed by repeatedly stacking multiple single cells in the order of bipolar plate-positive carbon felt-ion conduction membrane-negative carbon felt-bipolar plate and pressing them together with end plates and fastening bolts, and is used to carry out electrochemical reactions to achieve energy conversion. The electrolyte storage tank unit is used to store the electrolyte participating in the electrochemical reaction. The circulation pump group is used to drive the electrolyte to circulate between the storage tank and the stack. The piping system is used to connect the storage tank, the pump group, and the stack to form a closed circulation loop. The heat exchange unit is used to exchange heat with an external liquid cooling system to regulate the electrolyte temperature.

[0007] Furthermore, the multiphysics sensing network includes: an ultraviolet-visible spectrometer installed on the positive and negative electrode electrolyte circulation pipelines for real-time acquisition of the concentration and proportion of vanadium ions in various valence states; an online density meter installed on the main electrolyte pipeline for monitoring changes in the total vanadium concentration of the electrolyte; temperature sensors installed on the electrolyte inlet pipeline, outlet pipeline, and inside the storage tank of the fuel cell stack for monitoring the electrolyte temperature distribution; a distributed fiber optic grating sensor array embedded inside the fuel cell stack for real-time acquisition of temperature and stress field distribution data at the interface between the bipolar plates, carbon felt electrodes, and ion membrane inside the fuel cell stack; and an environmental temperature and humidity sensor installed outside the system compartment and an industrial communication interface for communication with the upper-level control system.

[0008] Furthermore, the electrolyte state adaptive module includes a prediction model constructed based on a physical information neural network. The input layer of the prediction model includes time variables, number of cycles or cumulative charge / discharge, current electrolyte temperature, current concentration of vanadium ions in each valence state, and current total vanadium concentration in the electrolyte. The output layer includes the rate of change of concentration of each valence state after a future period, probability of vanadium ion deposition, electrolyte viscosity, and electrolyte conductivity. The loss function of the prediction model includes data-driven terms and physical constraint terms. The physical constraint terms include residuals of the Nernst-Planck equation and the Butler-Folmer equation, which are used to embed the laws of ion diffusion and migration and electrode reaction kinetics as constraints into the model training process. The electrolyte state adaptive module determines whether to trigger temperature regulation and circulation pump flow rate regulation based on the output of the prediction model.

[0009] Furthermore, the loss function L for:

[0010] in, This is a data-driven term used to measure the deviation between model predictions and sensor measurements. These are physical constraint terms used to measure the degree to which the model output deviates from the physical equations; This is a regularization term used to prevent overfitting; and These are the weighting coefficients used to balance the contributions of each loss term; Data-driven items The weighted mean square error method is used:

[0011] in, N samples This represents the number of samples in the training batch. N output Output the number of variables for the model; w j For the first j The weighting coefficients of each output variable are used to balance the contribution of outputs with different dimensions to the loss function; y i,j For the first i The first sample j The sensor measured value or experimental calibration value of each output variable. This corresponds to the network prediction value; For the first j The standard deviation of each output variable; Data-driven items ,in and These are the weighting coefficients, and ; For the residuals of the Nernst-Planck equation, Let be the residual of the Butler-Folmer equation, and have

[0012] in, The number of configuration points refers to the number of computation points randomly or uniformly selected within the spatiotemporal domain. The model predicts the first i At the configuration point, the first k The concentration of seed ions, For the first i The potential at each configuration point For the first i The reaction source item at each configuration point For the first k The diffusion coefficient of vanadium ions in different valence states. For the first k The charge number of vanadium ions in each valence state. It is Faraday's constant. Let be the ideal gas constant. Absolute temperature t For time variables, x This refers to the number of cycles or the cumulative charge / discharge amount.

[0013] in, For the first i Current density at each configuration point For the first i Overpotential at each configuration point For the first i Temperature at each configuration point This represents the exchange current density calculated based on concentration and temperature. and These are the charge transfer coefficients for the anode and cathode, respectively.

[0014] Furthermore, the temperature regulation command is generated as follows: when the predicted deviation of the proportion of vanadium ions in each valence state is ≥10% or the vanadium ion deposition probability is ≥5%, the optimal operating temperature range under the current working conditions is determined. In a low-temperature environment, the electrolyte is heated to the range of 10-15℃, and in a high-temperature environment, the electrolyte is cooled to the range of 35-40℃. The regulation command is then sent to the heat exchange unit through the PID controller to guide the electrolyte circulation temperature to the target range.

[0015] Furthermore, the method for generating the circulating pump flow rate adjustment command is as follows: when the electrolyte viscosity is predicted to exceed the reference value by more than 20%, or when the model predicts or the electrolyte conductivity measured by the online conductivity sensor decreases by more than 15% relative to the reference conductivity, the fuzzy PID control algorithm is activated to dynamically adjust the circulating pump speed, increasing the circulating pump flow rate from the rated value by 10-20% to enhance convective heat transfer and ion transport. When the viscosity returns to normal and the conductivity recovers to more than 95% of the reference value, the circulating pump speed is gradually reduced to allow the system to return to the rated operating state.

[0016] Furthermore, the specific method by which the self-balancing module of the fuel cell stack thermal field adjusts the electrolyte flow distribution of each branch channel is as follows: Electric regulating valves are installed at the inlets of multiple parallel branch channels inside the fuel cell stack, and temperature sensors are installed in the corresponding fuel cell stack areas of each branch channel. The self-balancing module of the fuel cell stack thermal field sets an independent PID controller for each branch channel, compares the measured temperature of the corresponding area with the target operating temperature of that area, calculates the temperature deviation, and calculates the adjustment amount of the valve opening based on this deviation using an incremental PID algorithm. When the measured temperature is higher than the target temperature, the opening of the corresponding branch channel valve is increased to increase the electrolyte flow and enhance heat dissipation; when the measured temperature is lower than the target temperature, the opening of the corresponding branch channel valve is decreased to reduce the electrolyte flow and avoid excessive heat dissipation.

[0017] Furthermore, for any branch channel i Its corresponding first k The increment of valve opening at each sampling time and update volume They are represented as follows:

[0018]

[0019] in, For the first k- Valve opening at one sampling time, , and The first k , k -1、 k -Temperature deviation values ​​at 2 sampling times The sampling period is , and These are the corresponding proportional coefficient, integral coefficient, and differential coefficient, respectively.

[0020] Furthermore, the specific method by which the fuel cell stack thermal field self-balancing module adjusts the working state of the external liquid cooling system is as follows: the temperature values ​​of all temperature sensor measuring points inside the fuel cell stack are averaged to obtain the overall average temperature of the fuel cell stack, and this average temperature is compared with the overall target working temperature to calculate the overall temperature deviation; when the external liquid cooling system uses a variable frequency speed-controlled coolant circulation pump, the PID controller calculates the speed adjustment of the coolant circulation pump based on the overall temperature deviation. When the overall average temperature of the fuel cell stack is higher than the overall target temperature, the circulation pump speed is increased to increase the coolant flow and enhance the heat dissipation effect; when the overall average temperature of the fuel cell stack is lower than the overall target temperature, the circulation pump speed is decreased to reduce the coolant flow and reduce heat loss; when the external liquid cooling system uses air cooling, when the overall average temperature of the fuel cell stack exceeds the set upper limit, the cooling fan is started or the number of fans running is increased; when the overall average temperature of the fuel cell stack is lower than the set lower limit, the cooling fan is stopped or the number of fans running is reduced.

[0021] Compared with the prior art, the present invention has at least the following beneficial effects: At the state perception level, this invention overcomes the limitations of traditional technologies that only place temperature measurement points at the inlet and outlet of the fuel cell stack by embedding a distributed fiber Bragg grating sensor array inside the stack. This allows for real-time acquisition of temperature and stress distribution in different regions within the stack, accurately determining the location and severity of local hotspots, and providing a reliable sensing foundation for subsequent refined thermal field regulation. Simultaneously, by deploying an ultraviolet-visible spectrometer and an online density meter on the electrolyte circulation pipeline, the system can monitor the concentration of vanadium ions in various valence states, the total vanadium concentration, and their proportions in real time, promptly detecting valence imbalances and precipitation risks.

[0022] In terms of electrolyte state management, this invention constructs a PINN-based predictive model, incorporating physical laws such as vanadium ion diffusion kinetics and electrode reaction kinetics into the model training. This enables the system to anticipate potential valence imbalances or precipitation risks before obvious electrolyte anomalies appear, allowing for early intervention. Based on model predictions, the system actively adjusts the electrolyte temperature to a suitable operating range. In low-temperature environments, the temperature is increased to 10-15°C to reduce viscosity and suppress precipitation; in high-temperature environments, the temperature is decreased to 35-40°C to prevent thermal precipitation, achieving a shift from passive response to proactive prevention. Furthermore, based on model-predicted viscosity changes and measured data from online conductivity sensors, the system dynamically adjusts the circulation pump speed. When viscosity increases or conductivity decreases, the flow rate is increased by 10-20% to enhance mass transfer; under normal operating conditions, the pump speed is reduced to decrease energy consumption, achieving an effective balance between enhanced mass transfer and energy-saving operation.

[0023] Regarding the thermal field equalization of the fuel cell stack, this invention employs multiple temperature sensors arranged within the stack to monitor temperature differences in various regions. An independent electrically controlled regulating valve is installed for each branch flow channel. Based on the deviation between the measured temperature and the target temperature of the corresponding region, the electrolyte flow rate entering that channel is precisely adjusted. When the temperature in a certain region is too high, the valve opening of that branch flow channel is increased to enhance heat dissipation; when the temperature is too low, the valve opening is decreased to avoid excessive heat dissipation. This refined flow distribution effectively reduces the temperature difference between different regions within the stack, eliminates localized hot spots, thereby slowing down the oxidation and corrosion of the carbon felt electrodes and the aging process of the ion exchange membrane, extending the overall service life of the stack. Simultaneously, an external liquid cooling system regulates the overall temperature of the stack, forming a two-level thermal management architecture from global to local levels. This architecture can cope with overall heat load fluctuations caused by drastic changes in ambient temperature and solve the problem of localized temperature differences caused by uneven flow field distribution within the stack.

[0024] Through the above improvements, the vanadium redox flow storage system provided by this invention can operate stably in a wide temperature range of -30 to 50°C. The electrolyte will not be affected by freezing at low temperatures or precipitation at high temperatures, resulting in a more uniform temperature distribution inside the stack, effective suppression of local hot spots, extended stack cycle life, optimized energy consumption of the electrolyte circulation system, and significantly improved overall system reliability and economy. Attached Figure Description

[0025] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0026] Figure 1 This is a schematic diagram of the system framework provided in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention.

[0028] This embodiment provides a vanadium redox flow energy storage system based on self-organizing adaptation in a wide temperature environment, such as... Figure 1As shown, the system mainly includes a vanadium redox flow battery module, a multiphysics sensing network, and a full-temperature-range self-organizing adaptive controller. The full-temperature-range self-organizing adaptive controller is the core of the system; it is not a simple logic control unit, but an embedded intelligent agent integrating an electrolyte state adaptive module and a stack thermal field self-balancing module. The modules interact and make collaborative decisions through an internal high-speed bus.

[0029] I. Vanadium Redox Flow Battery Module The vanadium redox flow battery module adopts a modular design, consisting of several key components such as the stack unit, electrolyte storage tank unit, circulating pump group, piping system and heat exchange unit. These components work closely together through standardized interfaces and layout to form a complete and functionally independent energy conversion and storage unit.

[0030] As the core energy conversion component of the module, the fuel cell stack unit is an integral unit composed of dozens to hundreds of individual cells stacked and secured in a specific order. Each individual cell contains basic functional components such as bipolar plates, carbon felt electrodes, and ion-conducting membranes. The bipolar plates are typically made of carbon-plastic composite materials or graphite materials with excellent conductivity and corrosion resistance. They not only collect and conduct current, but their surfaces are also precisely designed with specific flow channels to guide the electrolyte to flow uniformly across the electrode surfaces. The carbon felt electrodes serve as the site of electrochemical reactions, and their porous structure provides ample surface area for vanadium ions to react. The ion-conducting membrane (usually a perfluorosulfonic acid membrane or a porous ion-conducting membrane) is sandwiched between the positive and negative electrodes, playing a crucial role in separating the electrolytes at the positive and negative electrodes while allowing specific ions (such as hydrogen ions) to pass through to form a complete circuit. During assembly, these components are repeatedly stacked in the order of bipolar plate - positive carbon felt - ion membrane - negative carbon felt - bipolar plate. Precise clamping force is applied through end plates and fastening bolts to ensure that the components fit tightly together, which ensures both airtightness and control of contact resistance.

[0031] In addition to the fuel cell stack unit, the module is equipped with an independent electrolyte storage tank unit for storing the active materials participating in the electrochemical reaction. Considering the strict separation of the positive and negative electrode electrolytes during charging and discharging, the system has two independent storage tanks, one for storing the positive electrode electrolyte (mainly containing tetravalent and pentavalent vanadium ions) and the other for storing the negative electrode electrolyte (mainly containing divalent and trivalent vanadium ions). These tanks are made of 316L stainless steel or high-density polyethylene, which have excellent corrosion resistance. The tank interiors are precision-machined or equipped with special flow-guiding structures to facilitate uniform mixing of the electrolyte and heat dispersion. The volume of the storage tanks directly determines the system's energy storage capacity; therefore, customized designs can be made according to the specific project duration requirements (e.g., 4 hours, 6 hours, or more than 8 hours).

[0032] To ensure continuous electrolyte circulation between the storage tank and the fuel cell stack, a circulation pump unit is integrated into the module. This pump unit employs a magnetically driven pump resistant to strong acid corrosion, connected to the motor via magnetic coupling rather than a mechanical shaft, thus completely eliminating the risk of electrolyte leakage. The selection and speed design of the circulation pump fully consider the matching of pipeline resistance and flow requirements, ensuring that under rated operating conditions, the electrolyte can flow through the internal channels of the fuel cell stack at a stable rate, promptly removing the heat generated by the reaction and continuously replenishing reactants to maintain the smooth progress of the electrochemical reaction. The corresponding piping system connects the storage tank, pump unit, and fuel cell stack, forming a closed circulation loop. The piping material is also selected from corrosion-resistant materials (such as PVDF or PTFE), and all connection points are made using welding or high-sealing flanges, clamps, etc. Pressure, temperature, and flow sensor interfaces are installed at key nodes for real-time monitoring of the circulation status.

[0033] In terms of thermal management, the module integrates a heat exchange unit, which centrally regulates the overall temperature of the electrolyte. This unit uses a plate heat exchanger, which is compact and highly efficient. The electrolyte side of the heat exchanger is connected to the main circulation loop, while the other side is connected to an external independent liquid cooling system. When the system operates at high temperatures or is in a high-power charging / discharging state, electrochemical reactions and ohmic resistance generate a large amount of heat. At this time, the external liquid cooling system will pass a coolant (such as an aqueous ethylene glycol solution) into the heat exchanger, where it exchanges heat with the electrolyte, carrying away the excess heat and dissipating it into the environment. Conversely, when starting the system at low temperatures, the external liquid cooling system can provide heat to preheat the electrolyte through the heat exchanger, allowing the electrolyte temperature to rise rapidly to the appropriate operating range, thereby avoiding the impact on system performance due to excessively high viscosity at low temperatures or vanadium ion precipitation.

[0034] II. Multiphysics Sensing Network The multiphysics sensing network is responsible for collecting system operating status data.

[0035] In terms of electrolyte state monitoring, this embodiment sets up independent online monitoring branches on the positive and negative electrolyte circulation pipelines to ensure that the respective states of the positive and negative electrolytes can be acquired in real time. Each monitoring branch is equipped with a flow cell made of corrosion-resistant transparent quartz material, with light windows on both sides to allow light transmission. On one side of the flow cell, a combined light source is connected via optical fiber. This light source integrates a deuterium lamp and a tungsten lamp, capable of stably outputting a continuous spectrum covering a wavelength range of 200 nm to 800 nm, a range sufficient to cover the characteristic absorption peaks of all vanadium ion valence states. On the other side of the flow cell, a high-resolution fiber optic spectrometer, such as the USB4000 model manufactured by Ocean Optics, is connected. This spectrometer can rapidly acquire absorbance spectral data after passing through the electrolyte at a sampling frequency of twice per second. During data processing, the system first normalizes the spectral data using the isoabsorption point at 600 nm to eliminate measurement errors caused by light intensity fluctuations and optical path contamination. Then, it extracts the characteristic absorbance values ​​of tetravalent vanadium ions at approximately 630 nm, pentavalent vanadium ions at approximately 760 nm, divalent vanadium ions at approximately 400 nm, and trivalent vanadium ions at approximately 450 nm. Substituting these absorbance values ​​into the Lambert-Beer law equations, pre-established through standard concentration solutions, allows for the real-time calculation of the concentrations of vanadium ions in each valence state in the electrolyte and their proportional relationships. This spectral analysis process is a standard technique in the field; specific optical component selection and calibration methods can be found in relevant standard literature or equipment manuals, and will not be elaborated upon here.

[0036] In addition to monitoring the vanadium ion valence state distribution, this embodiment also installs an online densitometer, such as the FDM series product from Endress+Hauser, on the main electrolyte pipeline. This densitometer operates based on the vibrating tube principle and has a measurement accuracy of ±0.0005 g / cm³, capable of reflecting subtle changes in the total vanadium concentration of the electrolyte in real time. Simultaneously, to comprehensively monitor the electrolyte temperature, this embodiment installs PT100 platinum resistance temperature sensors on the electrolyte inlet and outlet pipelines of the fuel cell unit, as well as inside the positive and negative electrode tanks. These sensors have a temperature measurement accuracy of ±0.1 degrees Celsius, and their measurement signals are connected to the analog acquisition module via a four-wire connection to eliminate measurement errors caused by wire resistance. Through temperature sensors distributed at different locations in the circulation loop, the system can monitor the temperature rise of the electrolyte before and after flowing through the fuel cell stack, as well as the temperature distribution within the tanks, in real time.

[0037] In terms of monitoring the internal condition of the fuel cell stack, this embodiment employs distributed fiber optic grating (FBG) sensing technology to achieve precise measurement of the internal temperature and stress distribution. During the assembly of the fuel cell stack, multiple FBG sensors are pre-embedded inside the stack. Twenty grating dots are fabricated on each fiber using ultraviolet laser etching technology. These grating dots have center wavelengths evenly distributed within the range of 1525 nm to 1565 nm, with a wavelength interval of approximately two nanometers between adjacent grating dots. This ensures that the reflected signals from each grating dot can be effectively distinguished in the frequency domain, thereby enabling simultaneous measurement of multiple measurement points using a single fiber. Specifically, in the implantation process, microgrooves approximately 0.2 mm deep and 0.3 mm wide are first fabricated on the surface of the bipolar plate using precision machining or laser etching. The orientation of these microgrooves is carefully designed so that the fiber can extend along the surface of the bipolar plate and pass through key areas at the junction of the carbon felt electrode and the ion exchange membrane. Subsequently, fiber Bragg grating sensors are carefully embedded into these microgrooves and fixed and encapsulated using epoxy resin with good thermal conductivity to ensure good thermal contact and mechanical transfer between the optical fibers and the bipolar plates. All fiber leads are converged outside the stack and connected to a fiber Bragg grating demodulator, such as the Micron Optics SM130-700 model. This demodulator can scan the center wavelength drift of each grating point in real time at a sampling frequency of 1000 times per second, achieving a wavelength resolution on the order of picometers. Since the center wavelength drift of the grating is linearly related to the changes in temperature and strain, typically with a temperature sensitivity of approximately 10 picometers per degree Celsius and a strain sensitivity of approximately 1.2 picometers per microstrain, the wavelength drift data measured by the demodulator can be used to deduce the real-time temperature and strain values ​​at each measuring point. Based on this, spatial interpolation algorithms are used to reconstruct the data from these discrete measuring points into two-dimensional or three-dimensional temperature and stress field distribution maps of the entire stack, thus visually revealing whether there are localized overheating areas or stress concentration phenomena within the stack.

[0038] Regarding environmental and electrical information acquisition, this embodiment installs an industrial-grade temperature and humidity sensor on the exterior of the energy storage system cabin to collect ambient temperature and humidity data in real time. This data will serve as an important reference for the system to assess external climate conditions and formulate operational strategies. Simultaneously, the system interacts with the upper-level control system via a standard industrial communication interface. For example, key electrical information such as system charge / discharge power commands, system state of charge (SOC), and grid dispatch requirements can be periodically read from the energy management system (EMS) or battery management system (BMS) via the controller area network bus (CAN bus) or Modbus-TCP industrial Ethernet protocol.

[0039] III. Full-temperature-range self-organizing adaptive controller As the core of the system, the full-temperature-range self-organizing adaptive controller is implemented using an embedded industrial computer (such as an industrial control board based on an ARM Cortex-A series processor), with a built-in Linux real-time operating system, and deploys an electrolyte state adaptive module and a stack thermal field self-balancing module.

[0040] 1. Electrolyte state adaptive module As a core component of the full-temperature-range self-organizing adaptive controller, the electrolyte state adaptive module's main task is to predict the future state evolution of the electrolyte based on electrolyte-related data collected by the multi-physics sensing network, and generate active intervention control commands accordingly.

[0041] 1.1 Constructing a state prediction neural network model We construct a prediction model based on Physics-Informed Neural Networks (PINN). PINN is a deep learning model that embeds physical laws (usually described in the form of partial differential equations) as constraints into the neural network training process. By designing a loss function that includes the residuals of physical equations, it enables the model to conform to basic physical laws while fitting data, thereby achieving efficient and reliable modeling and solving of complex scientific computing problems with scarce data.

[0042] In this embodiment, the prediction model includes an input layer, several hidden layers, and an output layer.

[0043] The input layer contains the following eight types of parameters: (1) The time variable t represents the running history of the system.

[0044] (2) Simplified representation of spatial coordinates: Considering that different locations within the vanadium redox flow battery exhibit state differences due to variations in electrolyte flow paths and electrochemical reaction rates during operation, but to reduce model complexity while maintaining prediction accuracy to meet real-time control requirements, a complete three-dimensional spatial coordinate system is not used here. Instead, the spatial dimension is simplified to the number of cycles or the cumulative charge / discharge amount. The number of cycles reflects the aging process of the electrode materials, while the cumulative charge / discharge amount reflects the total amount of reaction experienced by the electrolyte. These two parameters indirectly characterize the state distribution differences in different regions within the stack caused by the accumulation of reaction processes.

[0045] (3) The current electrolyte temperature T directly determines the ion diffusion coefficient and the reaction rate constant.

[0046] (4) The current concentration of divalent vanadium is denoted as C_V2.

[0047] (5) The current concentration of trivalent vanadium is denoted as C_V3.

[0048] (6) The current concentration of tetravalent vanadium is denoted as C_V4.

[0049] (7) The current concentration of pentavalent vanadium is denoted as C_V5.

[0050] (8) The current total vanadium concentration in the electrolyte, denoted as C_total, reflects the total amount of active substances in the electrolyte.

[0051] The output layer is designed to predict six key parameters after a future time interval Δt, specifically including: (1) The concentration change rate of vanadium ions in each valence state, i.e. the numerical value of the concentration change per unit time, is used to characterize the dynamic evolution trend of the valence state distribution. (2) Vanadium ion precipitation probability P_precip, used to warn of the risk of solid precipitation in electrolyte due to temperature change or concentration supersaturation; (3) Electrolyte viscosity μ is a key physical property parameter that affects flow resistance and mass transfer efficiency; (4) The electrolyte conductivity σ determines the electrolyte’s ion transport capacity.

[0052] The number of hidden layers, the number of neurons per layer, and the choice of activation functions need to be optimized based on the scale and complexity of the training data. A typical configuration involves 4 to 6 hidden layers, each containing 64 to 128 neurons, with hyperbolic tangent or modified linear unit (MLU) activation functions. The data received by the input layer is first normalized, scaling all parameters to near zero to accelerate convergence. The output layer uses different activation functions depending on the properties of the target being predicted. A linear activation function can be used for the concentration change rate to allow for positive and negative outputs, while the precipitation probability uses the sigmoid function to limit the output to between zero and one. Viscosity and conductivity are measured using the Softplus function to ensure positive outputs.

[0053] The construction of the loss function is the key difference between PINN and traditional neural networks. In this embodiment, the loss function is a weighted sum of data-driven terms, physical constraint terms, and regularization terms, i.e.

[0054] in, This is a data-driven term used to measure the deviation between model predictions and sensor measurements. These are physical constraint terms used to measure the degree to which the model output deviates from the physical equations; This is a regularization term used to prevent overfitting; and The typical value ranges for the weighting coefficients used to balance the contributions of each loss term are as follows: , .

[0055] Data-driven items We employ a weighted mean square error approach, considering the differences in the dimensions of different output variables, and introduce an adaptive weighting factor.

[0056] in, N samples This represents the number of samples in the training batch. N output In this embodiment, the number of output variables for the model is specified. N output = 7, corresponding to the concentration change rate of each valence state (4), precipitation probability, viscosity, and conductivity; w j For the first j The weighting coefficients of each output variable are used to balance the contribution of outputs of different dimensions to the loss function. The values ​​are determined experimentally (the weight related to the rate of change of concentration is set to 1.0, the weight related to viscosity and conductivity is set to 0.5, and the weight related to precipitation probability is set to 2.0). y i,j For the first i The first sample j The sensor measured value or experimental calibration value of each output variable. This corresponds to the network prediction value; For the first j The standard deviations of the output variables are:

[0057] N total The total number of samples in the training set. For the first j The mean of each output variable in the training set.

[0058] Physical constraints It consists of two sub-terms, corresponding to the residuals of the Nernst-Planck equation and the Butler-Volmer equation, respectively:

[0059] and These are the weighting coefficients, and ; This represents the residual of the Nernst-Planck equation, which describes the diffusion and migration behavior of ions under the combined influence of concentration and electric field gradients. For vanadium ions in the electrolyte of an all-vanadium redox flow battery, its one-dimensional simplified form is:

[0060] in, For the first k The concentration of vanadium ions in different valence states. t For time, x A simplified representation of spatial coordinates. For the first k The diffusion coefficient of vanadium ions in different valence states. For the first k The charge number of vanadium ions in each valence state. It is Faraday's constant. Let be the ideal gas constant. Absolute temperature For electric potential, For the first k The reaction source term for vanadium ions in various valence states.

[0061] In the PINN framework, the residuals of the Nernst-Planck equations are calculated via automatic differentiation, and their discrete form is as follows:

[0062] in, The number of configuration points refers to the number of computation points randomly or uniformly selected within the spatiotemporal domain. The first prediction of the neural network i At the configuration point, the first k The concentration of seed ions, For the first i The potential at each configuration point For the first i The reaction source item at each configuration point.

[0063] This represents the residual of the Butler-Folmer equation, which describes the nonlinear relationship between electrode reaction current density and overpotential. For the vanadium ion redox reaction, its form is:

[0064] in, j The electrode reaction current density, For exchange current density, and These are the charge transfer coefficients for the anode and cathode, respectively. This is an overpotential. In the PINN framework, the residuals of the Butler-Folmer equation are calculated as follows:

[0065] in, For the first i Current density at each configuration point For the first i Overpotential at each configuration point For the firsti Temperature at each configuration point This represents the exchange current density calculated based on concentration and temperature.

[0066] Regularization term Use L2 regularization, that is ,in This represents a vector consisting of all trainable parameters of a neural network. This indicates that the sum of squares of the elements of the parameter vector is calculated.

[0067] The organization and preparation of training data are divided into two stages: offline pre-training and online fine-tuning. In the offline pre-training stage, a training dataset covering a wide range of operating conditions needs to be constructed. This dataset was obtained through laboratory testing. The testing equipment included a complete vanadium redox flow battery testing system capable of precisely controlling key parameters such as temperature, circulation flow rate, and charge / discharge rate. During testing, the system operated under different temperature conditions, covering a range of -30 to 50 degrees Celsius; different circulation flow rates were set, from 50% to 150% of the rated flow rate; and different charge / discharge rates were used, from 0.5 to 2. Under each combination of operating conditions, continuous operation and aging tests were conducted, and electrolyte samples were periodically collected for offline analysis to obtain the true values ​​of parameters such as valence concentration, viscosity, and conductivity. Simultaneously, online measurement data from the sensors were recorded. These experimental data, after cleaning and normalization, constituted the offline training set for learning the initial parameters of the model. The online fine-tuning stage was carried out after the system was actually deployed and running. As the system operates under real-world conditions for an extended period, new operational data will continuously be generated, including data on various environmental temperature changes, load fluctuations, and the state evolution of the electrolyte during its natural aging process. This new data possesses characteristics specific to real-world operating conditions and is crucial for improving the model's prediction accuracy in practical applications. This embodiment employs a data caching mechanism, periodically adding newly collected valid data samples to the training set and using a mini-batch gradient descent algorithm to incrementally update the model parameters, enabling continuous online learning. The frequency of online fine-tuning can be set based on available computing resources; for example, model parameter fine-tuning can be performed once a week, with each fine-tuning using only recently accumulated new data to ensure the model can promptly track the slow changes in the system's state.

[0068] 1.2 Control Command Generation Based on the output of the PINN model, the electrolyte state adaptive module executes the following control logic: (1) Generation of temperature regulation target value First, based on the PINN model's prediction of the electrolyte state evolution trend over the next hour, a determination is made as to whether active temperature intervention is necessary. This determination is based on two parallel triggering conditions: First, the deviation of the proportion of vanadium ions in each valence state is ≥10%, meaning that the deviation between the current valence state distribution and the ideal distribution exceeds the safety margin; second, the model predicts that the probability of vanadium ion deposition is ≥5%, which means that the electrolyte is about to enter the thermodynamically unstable region.

[0069] When any of the above conditions are met, the module immediately initiates the temperature regulation process to determine the optimal operating temperature range under the current conditions. In low-temperature environments, because the viscosity of the electrolyte increases exponentially with decreasing temperature, the diffusion coefficient and solubility of vanadium ions decrease significantly. In this case, it is recommended to raise the electrolyte temperature to the range of 10-15℃ to reduce viscosity, enhance ion transport capacity, and suppress the risk of low-temperature precipitation. In high-temperature environments, the thermal motion of vanadium ions intensifies, and the solubility decreases with increasing temperature, making thermal precipitation more likely. In this case, it is preferable to lower the electrolyte temperature to the range of 35-40℃ to maintain electrolyte stability.

[0070] After determining the target temperature range, the module sends a PID control command to the heat exchange unit through an incremental proportional-integral-derivative controller. The controller takes the deviation between the target temperature and the actual temperature as input, and outputs a control quantity after calculation through the proportional, integral, and derivative stages. This drives the regulating valve or heater of the external liquid cooling system to gradually guide and stabilize the electrolyte circulation temperature within the target range.

[0071] (2) Circulating pump flow control When the PINN model predicts that the electrolyte viscosity exceeds the reference value by more than 20%, it signifies a significant increase in flow resistance, potentially affecting the uniformity of electrolyte distribution within the stack. This leads to decreased electrolyte flowability, increased pumping energy consumption, and a decline in ion transport efficiency between the electrolyte bulk and the electrode surface. Conversely, when the model predicts or the online conductivity sensor measures an electrolyte conductivity decrease of more than 15% relative to the reference conductivity, it indicates a reduction in the concentration of freely moving ions or a weakening of ion migration capacity. This is typically due to temperature deviations from the optimal range, an imbalance in the proportion of vanadium ions in different valence states, or localized concentration polarization. The decrease in conductivity directly reflects a weakening of the electrolyte's ion transport capacity and is a key indicator of intensified concentration polarization.

[0072] To address the two abnormal operating conditions of increased viscosity and decreased conductivity, the module employs a fuzzy PID control algorithm to dynamically adjust the speed of the circulating pump. When the electrolyte viscosity increases sharply, the fuzzy controller prioritizes increasing the proportional gain for a rapid response, while simultaneously increasing the integral gain to eliminate steady-state errors. This increases the circulating pump speed by 10-20% from its rated value, enhancing convective heat transfer and ion transport rates. When conductivity decreases significantly, the controller similarly increases the pump speed, but the focus is on improving mass transfer conditions at the electrode surface by enhancing flow, thus alleviating concentration polarization caused by limited mass transfer. Once the viscosity returns to normal and the conductivity recovers to above 95% of the reference value, the controller gradually reduces the proportional and integral gains, lowering the pump speed to reduce unnecessary energy consumption and bringing the system back to its rated operating state. This control strategy ensures timely enhancement of circulation under limited mass transfer conditions while maintaining low pump power loss under normal operating conditions, achieving efficient operation of the electrolyte circulation system.

[0073] 2. Fuel cell stack thermal field self-balancing module The self-balancing module for the thermal field of the fuel cell stack is crucial for the stable operation of vanadium redox flow storage systems (VDRs) across a wide temperature range. Its main function is to address localized overheating caused by uneven current distribution and electrolyte flow during charging and discharging. During stack operation, slight differences in electrolyte flow rates between individual cells and varying electrochemical reaction rates in different regions can lead to the generation of excessive heat in certain areas, forming localized hotspots. If these hotspots are not controlled promptly and effectively, their long-term accumulation can accelerate the oxidation and corrosion of the carbon felt electrodes, shorten the lifespan of the ion-exchange membrane, and even pose safety hazards.

[0074] The self-balancing thermal field module of the fuel cell stack operates on a multi-point temperature sensing network deployed inside the stack. During stack assembly, multiple temperature sensors are pre-positioned at key locations. These sensors, typically thermocouples or platinum resistance temperature sensors, are characterized by fast response, high measurement accuracy, and good long-term stability. The sensors are installed on the surface of the bipolar plates, at the interface between the electrodes and the ion exchange membrane, and at the inlet and outlet of the stack, covering temperature changes at various typical locations within the stack. The leads of each sensor are led out to the outside of the stack via sealed connectors and connected to a dedicated temperature acquisition module. The temperature acquisition module reads the temperature values ​​at each measurement point at a fixed sampling frequency, performs analog-to-digital conversion, and uploads the data in real time to the self-balancing thermal field module. This allows the module to obtain a real-time image of the temperature distribution inside the stack, identifying which areas are too hot and which are too cold.

[0075] Based on the temperature data obtained at various measuring points inside the fuel cell stack, the stack's thermal field self-balancing module achieves thermal field balance control by adjusting the flow distribution of electrolyte in each branch channel. The vanadium redox flow battery stack has multiple parallel flow channels. The electrolyte enters from the main inlet pipe and is distributed to each branch channel, flowing through the corresponding single-cell area before converging into the main outlet pipe. Since the electrolyte carries away the heat generated by the electrochemical reaction during convection, a larger electrolyte flow rate through a certain area results in a more significant heat dissipation effect and a relatively lower temperature in that area. Based on this principle, the module installs an electrically adjustable valve at the inlet of each branch channel. The flow rate of electrolyte entering that channel is adjusted by controlling the valve opening. When the temperature in a certain area is too high, the module increases the opening of the corresponding branch channel valve, allowing more electrolyte to flow through that area, enhancing the heat dissipation effect and gradually lowering its temperature. Conversely, when the temperature in a certain area is too low, the module appropriately decreases the valve opening, reducing the electrolyte flow rate in that area and preventing excessive heat dissipation. Through this refined flow distribution adjustment, the module can effectively reduce the temperature difference between different areas inside the fuel cell stack, making the temperature of the entire fuel cell stack more uniform.

[0076] In addition to regulating the flow distribution of each branch channel, the fuel cell stack thermal self-balancing module also controls the overall temperature of the entire stack through an external liquid cooling system. The external liquid cooling system includes a circulating pump, a radiator or cooling tower, and connecting pipes. The coolant circulates within the system, exchanging heat with the electrolyte through a plate heat exchanger installed on the stack's inlet pipe. When the overall stack temperature is too high, the module instructs the liquid cooling system to increase the circulating pump speed or activate the cooling fan, enhancing the coolant's ability to remove heat and thus lowering the electrolyte's inlet temperature, resulting in a decrease in the overall stack temperature. Conversely, when the overall stack temperature is too low, the module reduces the cooling intensity to maintain appropriate electrolyte temperature. This overall temperature control, combined with the branch channel flow distribution adjustment, forms a two-tiered thermal management architecture from global to local, capable of handling overall heat load fluctuations caused by drastic changes in ambient temperature while eliminating localized temperature differences within the stack due to uneven flow field distribution.

[0077] In terms of control strategy, the fuel cell stack thermal field self-balancing module adopts the PID control algorithm, a widely used and reliable control method in the industrial control field. For the flow distribution regulation of branch channels, the module sets an independent PID controller for each branch channel. It compares the measured temperature at the corresponding measuring point with the target operating temperature of the area, calculates the temperature deviation, and then calculates the adjustment amount of the corresponding electric regulating valve opening through the PID algorithm based on the magnitude, cumulative value, and trend of the deviation. This gradually adjusts the valve opening to a suitable value, thereby guiding the temperature of the area towards the target value.

[0078] For any branch flow channel i The input to its PID controller is the temperature deviation e. i (t) is defined as the target operating temperature T in this region. target_i Compared with the actual measured temperature T i The difference between (t), i.e., e i (t) =T target_i - T i (t). When the measured temperature is lower than the target temperature, e i (t) being positive indicates that heat dissipation in this area needs to be reduced, and the controller will output a command to reduce the valve opening; when the measured temperature is higher than the target temperature, e i If (t) is negative, it indicates that heat dissipation in this area needs to be strengthened, and the controller will output a command to increase the valve opening.

[0079] The PID controller calculates the valve opening adjustment based on this deviation signal. u i (t), its calculation formula is:

[0080] in, Indicates the first i The opening adjustment of each branch flow channel electric regulating valve at the current moment, in percentage increment of the opening relative to the rated position. A positive value indicates an increase in opening, and a negative value indicates a decrease in opening. The proportional coefficient determines the strength of the controller's response to the current deviation. The larger the proportional coefficient, the more sensitive the valve is to temperature deviation. The larger the deviation, the larger the valve adjustment range. The function of the proportional element is to respond quickly to temperature changes and make the system quickly approach the target value. The integral coefficient determines the controller's response strength to the accumulated deviation. The integral element can eliminate steady-state errors that may occur during long-term system operation. For example, when the valve opening needs to be maintained at a specific position to stabilize the temperature, the integral element will continuously accumulate small deviations and gradually adjust the valve opening until the temperature accurately reaches the target value. The derivative coefficient determines the controller's response strength to the trend of deviation changes. The derivative element can make predictions in advance based on the rate of temperature change. When the temperature rises rapidly, the derivative element will output a larger adjustment amount in advance to suppress temperature overshoot. When the temperature drops rapidly, it will reduce the adjustment amount in advance to prevent excessive temperature drop. The main function of the derivative element is to improve the dynamic response characteristics of the system and reduce temperature fluctuations.

[0081] In practical digital control systems, the continuous PID algorithm described above requires discretization. Since the temperature acquisition module reads the temperature values ​​at each measuring point at a fixed sampling period Ts, the controller calculates the valve opening adjustment at each sampling time k (k=0,1,2,…). The discretized incremental PID algorithm expression is:

[0082] in, Indicates the first k The increment of valve opening at each sampling moment is the adjustment range that needs to be adjusted this time. , and The first k , k -1、 k -Temperature deviation values ​​at two sampling times. Using this formula, the controller calculates the required valve opening increment based on the change between the current and previous deviations, the cumulative amount of the current deviation, and the acceleration of the deviation change. The actual valve opening is updated as follows: ,in This represents the valve opening at the previous sampling time. This incremental PID algorithm features low computational complexity and safe execution; even if an error occurs during the calculation, the actual valve opening will not change drastically.

[0083] proportionality coefficient Integral coefficient and differential coefficients The value of is crucial to the control effect and needs to be obtained through on-site debugging and tuning. Common tuning methods include the critical proportional coefficient method and the decay curve method. Taking the critical proportional coefficient method as an example, first, the integral and derivative coefficients are set to zero, and only proportional control is retained. The proportional coefficient is gradually increased until the system exhibits constant amplitude oscillation. The critical proportional coefficient Ku and the critical oscillation period Tu are recorded at this point. Then, the PID parameters are calculated according to the empirical formula: Kp i =0.6Ku, Ki i =1.2Ku / Tu, Kd i =0.075Ku·Tu. Since the different locations of each branch flow channel within the fuel cell result in varying flow resistance and heat dissipation conditions, the PID parameters for each flow channel need to be determined through an independent tuning process to achieve optimal control performance.

[0084] For overall temperature control of the fuel cell stack, the module adopts a strategy similar to that used for branch flow channel regulation, but the control objectives differ. The goal of overall temperature control is to maintain the average temperature of all measuring points on the fuel cell stack within the set optimal operating temperature range, thereby creating favorable conditions for local regulation of each branch flow channel. The module first calculates the average temperature T of all measuring points on the fuel cell stack at the current moment.avg (k), and then compare this average value with the overall target temperature Ttarget_avg to obtain the overall temperature deviation e. avg (k) = Ttarget_avg - T avg (k).

[0085] The control variables for the external liquid cooling system are calculated using a separate PID controller. The controlled objects of the liquid cooling system include the speed of the coolant circulation pump and the start / stop or speed of the cooling fan. Depending on the system configuration, the control output may be an analog signal or a switching signal. For coolant circulation pumps using variable frequency speed control, the controller output adjustment variable u... pump (k) Calculated according to the discrete PID formula:

[0086] in This represents the increment of the coolant circulation pump speed that needs to be adjusted at the current sampling time, in Hertz or percentage speed. , , These are the proportional, integral, and derivative coefficients for overall temperature control. Their tuning method is similar to that of branch flow channel control. However, because liquid cooling systems typically have a large time constant and slow response speed, the values ​​of these coefficients will differ significantly from those in branch flow channel control. Generally, a smaller proportional coefficient is needed to avoid system oscillation, while the integral coefficient is appropriately increased to eliminate steady-state errors. The actual speed update method for the coolant circulation pump is as follows: The updated speed command is sent to the frequency converter via the analog output module or digital communication interface. The frequency converter adjusts the motor speed, thereby changing the circulation flow rate of the coolant in the liquid cooling system and regulating the heat transfer intensity. When the overall temperature of the fuel cell stack is too high, the controller outputs a command to increase the speed, thereby increasing the coolant flow rate and enhancing the heat dissipation effect; when the overall temperature is too low, the controller outputs a command to decrease the speed, thereby reducing the coolant flow rate and minimizing heat loss.

[0087] For air-cooled heat dissipation systems employing start-stop control or tiered control, overall temperature control is implemented using different methods. When the average temperature exceeds the set upper limit, the controller starts the cooling fans or increases the number of running fans; when the average temperature falls below the set lower limit, the controller stops the fans or reduces the number of running fans. Although this control method is not as precise as variable frequency speed control, it is simple in structure and lower in cost, making it suitable for applications where temperature control accuracy requirements are not extremely stringent.

[0088] The fuel cell stack thermal self-balancing module also performs status monitoring and fault early warning functions. The module continuously records the temperature change trends at each measuring point. If the temperature in a certain area deviates from the normal range for an extended period, or if the temperature gradient continues to widen, the module will issue an early warning, prompting maintenance personnel to pay attention to potential problems such as flow channel blockage, seal failure, or electrode aging within the fuel cell stack. Simultaneously, the module uploads real-time temperature distribution data and valve opening status to the upper-level monitoring system via a communication interface, providing data support for operational analysis and fault diagnosis. When the temperature at a measuring point exceeds a safe threshold, the module will immediately send an alarm signal to the system-level controller, requesting a reduction in charging and discharging power or the implementation of other protective measures to prevent irreversible damage to the fuel cell stack due to excessive temperature.

[0089] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; under the concept of the present invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the present invention as described above, which are not provided in detail for the sake of brevity; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A vanadium redox flow energy storage system based on self-organizing adaptation in a wide-temperature environment, characterized in that, include: Vanadium redox flow battery module; Multiphysics sensing networks are used to acquire multi-dimensional state information of system operation in real time; A full-temperature-range self-organizing adaptive controller is connected to the vanadium redox flow battery module and the multi-physics sensing network, respectively. The full-temperature-range self-organizing adaptive controller includes an electrolyte state adaptive module and a stack thermal field self-balancing module. The electrolyte state adaptive module is used to receive electrolyte state information collected by the multi-physics sensing network, predict the future state evolution trend of the electrolyte based on the physical information neural network model, and generate temperature adjustment command and circulation pump flow rate adjustment command according to the prediction results, and send them to the vanadium redox flow battery module respectively. The self-balancing module for the thermal field of the fuel cell stack is used to receive the internal temperature field information of the fuel cell stack collected by the multi-physics sensing network, and adjust the electrolyte flow distribution of each branch channel according to the deviation between the measured temperature and the target temperature of each region. At the same time, it adjusts the working state of the external liquid cooling system according to the overall temperature condition of the fuel cell stack.

2. The all-vanadium redox flow energy storage system as described in claim 1, characterized in that, The vanadium redox flow battery module includes a stack unit, an electrolyte storage tank unit, a circulating pump group, a piping system, and a heat exchange unit. The fuel cell stack unit is composed of multiple single cells stacked repeatedly in the order of bipolar plate-positive carbon felt-ion conduction membrane-negative carbon felt-bipolar plate and then pressed together with end plates and fastening bolts. It is used to carry out electrochemical reactions to achieve energy conversion. The electrolyte storage tank unit is used to store the electrolyte participating in the electrochemical reaction. The circulating pump group is used to drive the electrolyte to circulate between the storage tank and the fuel cell stack. The pipeline system is used to connect the storage tank, the pump group and the fuel cell stack to form a closed circulation loop. The heat exchange unit is used to exchange heat with an external liquid cooling system to regulate the electrolyte temperature.

3. The all-vanadium redox flow energy storage system as described in claim 2, characterized in that, The multiphysics sensing network includes: an ultraviolet-visible spectrometer installed on the positive and negative electrode electrolyte circulation pipelines for real-time acquisition of the concentration and proportion of vanadium ions in various valence states; an online density meter installed on the main electrolyte pipeline for monitoring changes in the total vanadium concentration of the electrolyte; temperature sensors installed on the electrolyte inlet pipeline, outlet pipeline, and inside the storage tank for monitoring the electrolyte temperature distribution; a distributed fiber optic grating sensor array embedded inside the stack for real-time acquisition of temperature and stress field distribution data at the interface between the bipolar plates, carbon felt electrodes, and ion membrane inside the stack; and an environmental temperature and humidity sensor installed outside the system compartment and an industrial communication interface for communication with the upper-level control system.

4. The all-vanadium redox flow storage system as described in claim 1, characterized in that, The electrolyte state adaptive module includes a prediction model built based on a physical information neural network. The input layer of the prediction model includes time variables, number of cycles or cumulative charge / discharge, current electrolyte temperature, current concentration of vanadium ions in each valence state, and current total vanadium concentration in the electrolyte. The output layer includes the rate of change of concentration of each valence state after a future period, probability of vanadium ion deposition, electrolyte viscosity, and electrolyte conductivity. The loss function of the prediction model includes a data-driven term and a physical constraint term. The physical constraint term includes the residuals of the Nernst-Planck equation and the Butler-Folmer equation, which are used to embed the laws of ion diffusion and migration and the laws of electrode reaction kinetics as constraints into the model training process. The electrolyte state adaptive module determines whether to trigger temperature adjustment and circulation pump flow rate adjustment based on the output of the prediction model.

5. The all-vanadium redox flow energy storage system as described in claim 4, characterized in that, The loss function L for: in, This is a data-driven term used to measure the deviation between model predictions and sensor measurements. These are physical constraint terms used to measure the degree to which the model output deviates from the physical equations; This is a regularization term used to prevent overfitting; and These are the weighting coefficients used to balance the contributions of each loss term; Data-driven items The weighted mean square error method is used: in, N samples This represents the number of samples in the training batch. N output Output the number of variables for the model; w j For the first j The weighting coefficients of each output variable are used to balance the contribution of outputs with different dimensions to the loss function; y i,j For the first i The first sample j The sensor measured value or experimental calibration value of each output variable. This corresponds to the network prediction value; For the first j The standard deviation of each output variable; Data-driven items ,in and These are the weighting coefficients, and ; For the residuals of the Nernst-Planck equation, Let be the residual of the Butler-Folmer equation, and have in, The number of configuration points refers to the number of computation points randomly or uniformly selected within the spatiotemporal domain. The model predicts the first i At the configuration point, the first k The concentration of seed ions, For the first i The potential at each configuration point For the first i The reaction source item at each configuration point, For the first k The diffusion coefficient of vanadium ions in different valence states. For the first k The charge number of vanadium ions in each valence state. It is Faraday's constant. Let be the ideal gas constant. Absolute temperature t For time variables, x This refers to the number of cycles or the cumulative charge / discharge amount. in, For the first i Current density at each configuration point For the first i Overpotential at each configuration point For the first i Temperature at each configuration point This represents the exchange current density calculated based on concentration and temperature. and These are the charge transfer coefficients for the anode and cathode, respectively.

6. The all-vanadium redox flow energy storage system as described in claim 4, characterized in that, The temperature regulation command is generated as follows: when the predicted deviation of the proportion of vanadium ions in each valence state is ≥10% or the probability of vanadium ion deposition is ≥5%, the optimal operating temperature range under the current working conditions is determined. In a low-temperature environment, the electrolyte is heated to the range of 10-15℃, and in a high-temperature environment, the electrolyte is cooled to the range of 35-40℃. The regulation command is then sent to the heat exchange unit through the PID controller to guide the electrolyte circulation temperature to the target range.

7. The all-vanadium redox flow energy storage system as described in claim 4, characterized in that, The method for generating the circulating pump flow rate adjustment command is as follows: when the electrolyte viscosity is predicted to exceed the reference value by more than 20%, or when the model predicts or the electrolyte conductivity measured by the online conductivity sensor decreases by more than 15% relative to the reference conductivity, the fuzzy PID control algorithm is activated to dynamically adjust the circulating pump speed, increasing the circulating pump flow rate from the rated value by 10-20% to enhance convective heat transfer and ion transport. When the viscosity returns to normal and the conductivity rises back to more than 95% of the reference value, the circulating pump speed is gradually reduced to bring the system back to the rated operating state.

8. The all-vanadium redox flow storage system as described in claim 1, characterized in that, The specific method by which the self-balancing module of the fuel cell thermal field adjusts the electrolyte flow distribution in each branch channel is as follows: Electric regulating valves are installed at the inlet of multiple parallel branch channels inside the fuel cell stack, and temperature sensors are installed in the fuel cell stack area corresponding to each branch channel. The fuel cell stack thermal field self-balancing module is equipped with an independent PID controller for each branch channel. The controller compares the actual temperature measured by the temperature sensor in the corresponding area with the target operating temperature of the area to calculate the temperature deviation. Based on the deviation, the valve opening adjustment is calculated using an incremental PID algorithm. When the actual temperature is higher than the target temperature, the valve opening of the corresponding branch channel is increased to increase the electrolyte flow and enhance heat dissipation. When the actual temperature is lower than the target temperature, the valve opening of the corresponding branch channel is decreased to reduce the electrolyte flow and avoid excessive heat dissipation.

9. The all-vanadium redox flow energy storage system as described in claim 8, characterized in that, For any branch flow channel i Its corresponding first k The increment of valve opening at each sampling time and update volume They are represented as follows: in, For the first k- Valve opening at one sampling time, , and The first k , k -1、 k -Temperature deviation values ​​at 2 sampling times The sampling period is , and These are the corresponding proportional coefficient, integral coefficient, and differential coefficient, respectively.

10. The all-vanadium redox flow energy storage system as described in claim 1, characterized in that, The specific method by which the self-balancing module of the fuel cell thermal field adjusts the working state of the external liquid cooling system is as follows: The average temperature of the fuel cell stack is calculated by averaging the temperature values ​​of all temperature sensor points inside the stack. This average temperature is then compared with the overall target operating temperature to calculate the overall temperature deviation. For external liquid cooling systems using variable frequency speed-controlled coolant circulation pumps, the PID controller calculates the pump speed adjustment based on the overall temperature deviation. When the overall average temperature of the fuel cell stack is higher than the overall target temperature, the pump speed is increased to improve coolant flow and enhance heat dissipation. When the overall average temperature of the fuel cell stack is lower than the overall target temperature, the pump speed is decreased to reduce coolant flow and minimize heat loss. For external liquid cooling systems using air cooling, the cooling fans are activated or the number of fans is increased when the overall average temperature of the fuel cell stack exceeds the set upper limit. When the overall average temperature of the fuel cell stack is lower than the set lower limit, the cooling fans are stopped or the number of fans is reduced.

Images