Energy management and safety protection cooperation method for liquid cooling industrial and commercial energy storage system

Abstract

The invention discloses an energy management and safety protection cooperation method for a liquid cooling industrial and commercial energy storage system, and particularly relates to the technical field of energy storage system management. A battery electrochemical model, a heat distribution diagram, temperature gradient data and electrical parameters are used as input, and a battery temperature change trend curve is output; a liquid cooling control strategy is set according to the prediction result; fusing the temperature gradient abnormal parameters, the temperature trend risk and the multi-modal environment data abnormal parameters, starting a fire risk assessment model, predicting the fire probability and position, calculating a fire risk coefficient, generating a fire risk report and setting safety protection measures; a multi-objective optimization mathematical model is constructed based on the energy efficiency ratio, the full life cycle income and the battery health degree, energy storage operation data and power grid requirements are combined, a Pareto optimal solution set is generated by adopting a non-dominated sorting genetic algorithm, and a charging and discharging strategy and liquid cooling parameters are optimized; the liquid cooling pipeline layout is optimized through reinforcement learning, and the problem that the battery temperature cannot be effectively managed is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage system management, and more specifically, to a collaborative method for energy management and safety protection of a liquid-cooled industrial and commercial energy storage system. Background Art

[0002] As an efficient thermal management means, liquid cooling technology has been widely applied in industrial and commercial energy storage systems. By circulating the coolant to absorb the heat generated during the operation of the battery and export it, the battery can be maintained within a safe temperature range. However, the existing liquid cooling technology still faces many challenges in practical applications. Especially in large-scale energy storage scenarios, the electrochemical reactions and thermal behaviors of the batteries show high complexity and dynamics, resulting in problems such as uneven temperature distribution and increased risk of thermal runaway. Traditional liquid cooling systems usually rely on static or simple control strategies, such as fixed coolant flow rate and preset temperature thresholds, and are difficult to adapt to the dynamic thermal management requirements brought about by battery load fluctuations or environmental condition changes. This lag and low adaptability lead to insufficient cooling efficiency and decreased temperature control accuracy, not only reducing the energy efficiency of the energy storage system but also potentially hiding safety hazards, such as local overheating or even fire risks.

[0003] In addition, there are obvious deficiencies in the coordination between energy management and safety protection in the existing liquid cooling control systems. On the one hand, traditional methods lack the ability to predict the changing trend of battery temperature, often making passive adjustments only based on real-time data, and unable to prospectively optimize the cooling strategy, resulting in uneven distribution or over-cooling of cooling resources, and thus affecting the full life cycle benefits of the system and the battery health. On the other hand, the existing technologies mainly monitor the fire risk based on a single index (such as temperature exceeding the standard), and fail to effectively integrate multi-modal environmental data (such as vibration, gas concentration, etc.) for comprehensive analysis, resulting in insufficient accuracy of fire prediction and pertinence of protection measures. In high-load or abnormal working conditions, this single thermal management and risk control method is difficult to meet the requirements of modern industrial and commercial energy storage systems for efficient, safe, and intelligent operation. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a collaborative method for energy management and safety protection of a liquid-cooled industrial and commercial energy storage system to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solution: A collaborative method for energy management and safety protection of a liquid-cooled industrial and commercial energy storage system, comprising the following steps:

[0006] Simulate the electrochemical reaction and thermal behavior of the battery to generate a preset thermal distribution map and heat transfer parameters; collect the temperature, voltage, current data of the battery and multi-modal environmental data in real time through a sensor network, filter the noise by an edge computing device and calculate the temperature gradient, and output the real-time temperature distribution matrix and temperature gradient data;

[0007] Build a temperature trend prediction model using a long short-term memory network, input the preset thermal distribution map, heat transfer parameters, real-time temperature distribution matrix and electrical parameters, convert the spatial data into a time series format adapted to LSTM, output the battery temperature change trend curve, and set the liquid cooling control strategy based on the battery temperature change trend curve; dynamically adjust the coolant flow rate and temperature of the liquid cooling system;

[0008] Fuse the temperature gradient anomaly parameter, temperature trend risk and multi-modal environmental data anomaly parameter, start the fire risk assessment model, predict the fire probability and location, calculate the fire risk coefficient, generate a fire risk report and set safety protection measures;

[0009] Construct a multi-objective optimization mathematical model based on the energy efficiency ratio, full life cycle benefit and battery health, combine the energy storage operation data with the power grid demand, use the non-dominated sorting genetic algorithm to generate the Pareto optimal solution set, optimize the charge and discharge strategy and liquid cooling parameters, and introduce the Bayesian optimization technology to dynamically adjust the algorithm parameters; based on the real-time temperature distribution matrix and temperature change trend curve, determine continuous temperature anomalies (such as when the fire risk coefficient continuously exceeds the preset value, it is determined as continuous temperature anomaly); if the temperature is continuously abnormal, use reinforcement learning to optimize the liquid cooling pipeline layout, iteratively adjust the pipe diameter and flow distribution to ensure cooling uniformity and energy efficiency.

[0010] Preferably, based on the battery temperature change trend curve, adaptively adjust the acquisition frequency of battery data.

[0011] Preferably, the acquisition method of the fire risk assessment model is: construct a multi-modal environmental data set, including temperature data, vibration data, sound data and gas concentration data; extract features according to the multi-modal environmental information, at least including the temperature change speed, temperature acceleration, gas anomaly degree; build an initial fire risk assessment model based on a long short-term memory network, train the fire risk assessment model based on the training set, and use the cross-validation method to divide the historical data into a training set and a test set; find the optimal hyperparameter combination through methods such as grid search or random search, and conduct cross-validation and performance evaluation to obtain the trained fire risk assessment model.

[0012] Preferably, the method further includes a fire risk management step, including:

[0013] Environmental data collection: The sensor network also collects multi-environmental data such as the vibration, sound, and gas concentration (smoke sensor, gas concentration sensor) of the battery, and fuses them to obtain multi-modal environmental data;

[0014] Abnormality discrimination: Based on the temperature gradient abnormality parameter, temperature trend risk, and multi-modal environmental data abnormality parameter, a fire risk assessment model is triggered. The temperature trend risk refers to the potential overheating risk of the battery temperature within a certain period in the future based on the predicted temperature change trend; the multi-modal environmental data abnormality parameter refers to the degree to which the vibration, sound, and gas concentration exceed the preset range;

[0015] Fire risk assessment: Input the fused multi-modal environmental information into the fire risk assessment model, predict the probability and potential location of a fire, calculate the fire risk coefficient. The probability of a fire, potential location, and fire risk coefficient constitute a fire risk report, and safety protection measures are set based on the fire risk report; intelligent response (such as local isolation or fire extinguishing agent spraying) is automatically triggered.

[0016] Preferably, the response strategy is dynamically optimized to ensure both resource utilization efficiency and safety. The target monitored battery is divided into several regions, and the maintenance priority is determined based on the fire risk coefficient. The acquisition method of the fire risk coefficient is as follows:

[0017] Obtain the fire probability P fire , which is the fire occurrence probability predicted by the LSTM model within the future time, and the value range is [0, 1];

[0018] Combine the temperature trend risk T risk , multi-modal environmental data abnormality parameter H y and the fire probability, and obtain the fire risk coefficient C through the following formula risk =1 - exp[-k·(P fire ·(1 + T risk ))·(1 + H y )], where k represents a constant controlling the overall risk sensitivity, and the range is [1, 10].

[0019] Preferably, it also includes an operation optimization management step, including: constructing a multi-objective optimization mathematical model based on the energy efficiency ratio, full life cycle benefit, and battery health; using the multi-objective optimization mathematical model, combining the operation data of the energy storage system and the grid demand, to quantify the operation characteristics by constructing a multi-objective optimization mathematical model, initialize the non-dominated sorting genetic algorithm population to generate a preliminary strategy, and perform evolutionary calculation to generate a Pareto optimal solution set to achieve multi-objective collaborative optimization, optimize the charge and discharge strategy and the parameters of the liquid cooling system, and generate a Pareto optimal solution set with economy, reliability, and environmental protection; introduce Bayesian optimization technology to dynamically adjust the algorithm parameters to improve the optimization efficiency and global search ability.

[0020] Preferably, the operation optimization management step includes the following contents:

[0021] Step S31, constructing a multi-objective optimization mathematical model: extracting the power grid demand and key parameters of the energy storage system from the operation data, where the power grid demand data includes at least load and frequency regulation demand, and the key parameters include energy efficiency ratio, full life cycle benefit, and battery health. Jointly establish a multi-objective optimization mathematical model; output the multi-objective optimization mathematical model and the power grid demand data;

[0022] Step S32, initializing the non-dominated sorting genetic algorithm population: based on the multi-objective optimization mathematical model and the power grid demand data, use the non-dominated sorting genetic algorithm to initialize the population, and output the initial operation strategy set;

[0023] Step S33, evolving to generate a Pareto optimal solution set: based on the initial operation strategy set, use the non-dominated sorting genetic algorithm to perform evolutionary calculations, and output the Pareto optimal solution set;

[0024] Explanation: Identify the Pareto front layer, where the first layer is the current optimal strategy; through the tournament selection mechanism, screen the parent strategies from the population according to the crowding distance and domination level; use simulated binary crossover and polynomial mutation to adjust the decision variables to generate offspring strategies and expand the diversity of the solution space; iteratively execute the sorting, selection, and mutation processes until the convergence condition is met (such as the number of iterations reaches the preset value or the front layer is stable) to generate the Pareto optimal solution set; each set of solutions includes the optimized charge and discharge power and liquid cooling parameters, corresponding to the target values of energy efficiency ratio, full life cycle benefit, and battery health;

[0025] Step S34, Bayesian optimization adjustment: based on the Pareto optimal solution set, introduce Bayesian optimization technology to adjust the crossover probability and mutation index of the non-dominated sorting genetic algorithm to generate the final operation optimization control strategy to improve the performance of the solution set and the global optimization ability.

[0026] Preferably, it further includes the step of optimizing the layout of the liquid cooling pipeline:

[0027] If any one of the abnormal temperature gradient parameters, temperature trend risks, and abnormal multi-modal environment data parameters still exists in the operation optimization control strategy scenario, trigger the step of optimizing the layout of the liquid cooling pipeline;

[0028] Preferably, the process of optimizing the layout of the liquid cooling pipeline includes the following steps:

[0029] Obtain the battery temperature change trend curve and the real-time temperature distribution matrix, input the heat transfer parameters into the heat flow simulation model, and use reinforcement learning to iteratively adjust the pipe diameter, branches, and flow distribution. Based on the cooling uniformity, liquid cooling energy efficiency, and the highest temperature, set the reward target for pipeline optimization; generate the optimized pipeline design parameters to ensure both cooling uniformity and energy efficiency are considered;

[0030] Obtain the pipeline design parameters, adjust the liquid cooling pipeline layout, and verify the battery temperature change trend curve and the real-time temperature distribution matrix to ensure the reliability of pipeline optimization.

[0031] Preferably, the calculation method of the reward target is as follows:

[0032] Obtain the cooling uniformity, the higher the value, the better the uniformity; obtain the cooling efficiency Q remored Compared with the energy consumption E cooling Calculate the liquid cooling energy efficiency; calculate the reward target through the following formula;

[0033]

[0034] Among them, T i represents the temperature at a certain space-time position, w1, w2, and w3 respectively represent the weight coefficients of each item, T max is the highest temperature, T safe is the safety threshold, T ref is the reference temperature.

[0035] Preferably, the method further includes a fault location step:

[0036] If any one of the temperature gradient abnormal parameter, temperature trend risk, and multi-modal environment data abnormal parameter exists, trigger the fault location step;

[0037] Use the time series anomaly detection algorithm to monitor the operation data in real time and automatically identify abnormal behaviors; combine causal inference technology to analyze the causal relationship between the anomalies and components (such as batteries, liquid cooling units), and quickly locate the root cause; construct a fault knowledge graph, automatically classify abnormal patterns based on historical data, and recommend solutions;

[0038] Obtain the operation data of the energy storage system, apply a recurrent neural network based on the attention mechanism to analyze the time series characteristics, generate an abnormal behavior sequence to identify potential fault signals (such as abnormal temperature sudden increase or voltage sudden drop, etc.);

[0039] Obtain the abnormal behavior sequence, combine the structural causal model to infer the causal relationship between component faults and abnormal behavior sequences, and generate a fault causal diagram to locate the fault root cause;

[0040] Obtain the fault causal diagram, combine it with the graph neural network to fuse historical data, generate a fault diagnosis report to classify anomalies and recommend solutions.

[0041] Preferably, an abnormal behavior sequence is identified based on an abnormal behavior detection threshold function, and the calculation formula of the abnormal behavior detection threshold function is:

[0042]

[0043] where x t is the current data value, x pred represents the predicted value, x ref is the reference value, and λ represents the sensitivity factor.

[0044] The technical effects and advantages of the present invention:

[0045] (1) The method for coordinating energy management and safety protection of the liquid-cooled industrial and commercial energy storage system provided by the present invention combines battery electrochemical models, sensor networks, long short-term memory networks, fire risk assessment, and operation optimization management technologies to provide a comprehensive safety protection and energy efficiency optimization strategy; by real-time monitoring data such as battery temperature, current, and voltage, predicting the temperature change trend, and dynamically adjusting the control strategy of the liquid-cooled system, it ensures that the energy storage system avoids safety risks such as overheating while operating efficiently. In addition, the fire risk assessment model combines multi-modal environmental data such as temperature, vibration, and gas concentration to trigger intelligent response measures in a timely manner, improving the safety of the system and effectively solving the deficiencies of traditional energy storage systems in energy efficiency management and safety protection, and enhancing the operation efficiency and safety guarantee ability of the system.

[0046] (2) The method for coordinating energy management and safety protection of the liquid-cooled industrial and commercial energy storage system provided by the present invention constructs a multi-objective optimization mathematical model based on the energy efficiency ratio, full-life cycle income, and battery health, combines energy storage operation data with grid demand, and uses the non-dominated sorting genetic algorithm to generate a Pareto optimal solution set, effectively optimizing the charge and discharge strategy and liquid-cooled parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is the temperature control flow chart based on the temperature change trend curve of the present invention.

[0048] Figure 2 is the temperature control flow chart based on the operation strategy optimization of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0049] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be completely conveyed to those skilled in the art.

[0050] At the same time, it should be understood that for the convenience of description, the dimensions of the various parts shown in the drawings are not drawn in actual proportional relationship.

[0051] The following description of at least one exemplary embodiment is actually merely illustrative and in no way limits the present application and its application or use.

[0052] Technologies, methods, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the said technologies, methods, and devices should be regarded as part of the specification.

[0053] Example 1, referring to Figure 1 the temperature control flow chart based on the temperature change trend curve, the present invention provides a method for coordinating energy management and safety protection of a liquid-cooled industrial and commercial energy storage system as shown in Figure 1 and includes the following steps:

[0054] Step S11: Use finite element analysis software combined with a battery electrochemical model to simulate the electrochemical reactions and thermal behaviors of the battery under different working conditions, analyze the temperature distribution and heat transfer characteristics, and generate a preset heat distribution map and heat transfer parameters;

[0055] Explanation: The heat distribution map shows the temperature distribution of the battery under different working conditions. By comparing the heat distribution map with the actually measured battery temperature data, the accuracy of the thermal coupling model is calibrated; heat transfer parameters such as thermal resistance and heat capacity are extracted from the heat distribution map, and the heat transfer parameters reflect the heat transfer characteristics of the battery;

[0056] In the embodiment of the present invention, finite element analysis software (such as COMSOL Multiphysics) is used to construct a three-dimensional model of the battery, and the mesh division accuracy is 0.5 - 1 mm; the battery electrochemical model simulates lithium ion diffusion and charge transfer based on the Butler-Volmer equation, and the input parameters include electrode material properties (such as lithium diffusion coefficient) and working conditions (such as charging, discharging); a heat distribution map (two-dimensional temperature field, unit: °C) and heat transfer parameters are simulated and generated; the heat distribution map calibrates the model by comparing with the actual temperature data, and controls the error (such as the error is controlled within ±2 °C);

[0057] Step S12: Real-time collect the temperature, voltage, and current data of the battery through the sensor network. Filter the noise through the edge computing device and calculate the temperature gradient, and output the real-time temperature distribution matrix and temperature gradient data.

[0058] Explanation: The temperature gradient data reflects the uniformity of the battery temperature distribution. By analyzing the temperature gradient data, potential local overheating areas (such as a certain part with a significantly higher temperature than other areas) can be identified, so as to take timely measures to avoid the risk of thermal runaway.

[0059] In the embodiment of the present invention, the sensor network is deployed with thermistors (accuracy ±0.1°C), voltage sensors (range 0 - 5V), and current sensors (range 0 - 200A), covering the positive electrode, negative electrode, and central area of the battery, with a sampling frequency of 1Hz; the edge computing device (such as RaspberryPi) uses Kalman filtering to remove noise and calculate the temperature gradient (unit: °C / m, formula: temperature difference between adjacent points divided by the distance); generate a real-time temperature distribution matrix (two-dimensional array, each element is a temperature value), temperature gradient data (vector form), and electrical parameters (voltage and current sequences), and the temperature gradient data is used to identify local overheating (such as gradient > 5°C / m).

[0060] Step S13: Build a temperature trend prediction model using a long short-term memory network. Input the preset thermal distribution map, heat transfer parameters, real-time temperature distribution matrix, and electrical parameters, convert the spatial data into a time series format adapted to LSTM, predict the battery temperature change trend curve in the future for a period of time (such as 5 - 30 minutes), and set the liquid cooling control strategy (such as the timing control instructions for coolant flow rate and temperature set value) based on the battery temperature change trend curve.

[0061] In the embodiment of the present invention, the long short-term memory network model is preferably adopted with a 3-layer convolutional structure (convolution kernel 3×3), and the input dimensions are the thermal distribution map (two-dimensional), real-time temperature distribution matrix (two-dimensional), temperature gradient data (one-dimensional vector), and heat transfer parameters (scalar); take measures according to the battery temperature change trend curve. For example, if the temperature rising rate > 0.5°C / min, set the liquid cooling control strategy.

[0062] In a possible embodiment, adaptively adjust the data acquisition frequency according to the battery temperature change trend curve.

[0063] Step S14: Dynamically adjust the coolant flow rate and temperature of the liquid cooling system according to the liquid cooling control strategy, and real-time monitor the battery temperature. If the battery temperature exceeds the expectation, give an alarm and feedback data to optimize the long short-term memory network model to achieve continuous improvement.

[0064] In an embodiment of the present invention, a programmable logic controller is used to receive control parameters and adjust the flow rate of the liquid cooling pump and the coolant temperature; the temperature sensor in step S12 is monitored in real time, and the temperature change within 5 - 30 minutes after adjustment is recorded. If the measured temperature exceeds the predicted value by ±2°C, it is regarded as exceeding the expectation, and the real-time temperature data is fed back to step S13 to update the weights of the CNN model (it is recommended to use the Adam optimizer with a learning rate of 0.001); the adjustment effect is evaluated through temperature stability, and the optimized model parameters are output to enhance the long-term thermal management and safety protection capabilities.

[0065] In the embodiment of the present invention that needs further explanation, the method further includes a fire risk management step, including:

[0066] Environmental data collection: The sensor network also collects multi-environmental data such as the vibration, sound, and gas concentration (smoke sensor, gas concentration sensor) of the battery, and fuses them to obtain multi-modal environmental data;

[0067] Explanation: The multi-modal environmental data (temperature, vibration, gas concentration) is fused at the feature level: the time-domain features of each modality are extracted (such as the temperature change rate, vibration spectrum energy, gas concentration gradient), and weighted fusion is performed using the attention mechanism. The weight coefficients are obtained by training with historical data;

[0068] Abnormality discrimination: A fire risk assessment model is triggered based on the temperature gradient anomaly parameter, temperature trend risk, and multi-modal environmental data anomaly parameter. The temperature trend risk refers to the potential overheating risk of the battery temperature in a future period based on the predicted temperature change trend; the multi-modal environmental data anomaly parameter refers to the degree to which the vibration, sound, and gas concentration exceed the preset range;

[0069] In a possible embodiment, the triggering condition of the fire risk assessment model is that the comprehensive value of the multi-modal anomaly parameters exceeds the threshold, or a single parameter exceeds the safety limit; for example, when the temperature gradient anomaly parameter Wy > 5°C / m, or the gas concentration anomaly exceeds the preset threshold, the fire risk assessment is immediately triggered;

[0070] Furthermore, the present invention does not specifically limit the triggering conditions of the fire risk assessment model, which can be any one or a combination of the temperature gradient anomaly parameter, temperature trend risk, and multi-modal environmental data;

[0071] For easy understanding, in this embodiment, the fire risk assessment model is triggered based on the temperature trend risk: The temperature trend risk T risk refers to the risk accumulation from the current moment t to a future time period t + Δt, and is calculated by the following formula:

[0072]

[0073] Among them, represents the rate of change of temperature with respect to time; represents the acceleration of temperature change, and β1 is an adjustment parameter used to amplify or weaken the influence of acceleration; σ pred represents the temperature prediction confidence interval, reflecting the credibility of the future trend; β2 represents the uncertainty weight, used to adjust the contribution of the prediction error to the risk;

[0074] When T risk exceeds the threshold, execute the fire risk management steps;

[0075] In a possible embodiment, a fire risk assessment model is triggered based on the abnormal parameters of multi-modal environmental data. Let H y represent the abnormal parameters of multi-modal environmental data, with a value range from 0 to 1. The closer the value is to 1, the further the multi-modal environmental data deviates from the normal;

[0076] Explanation: The quantization formula for the abnormal parameters of the multi-modal environmental data is as follows:

[0077] For each type of environmental data (vibration, sound, gas concentration), calculate its abnormality degree through the formula ; x i represents the current sensor measurement value (such as vibration acceleration, sound decibel, gas concentration), μ i represents the mean value of historical normal data, and σ i represents the standard deviation of historical normal data;

[0078] Compress the abnormality degree of each type of environmental data to the range from 0 to 1 through the Sigmoid function;

[0079] Allocate weights according to the contribution of each type of environmental data to the fire risk, calculate the weighted sum of the normalized abnormality degree and the weights, and obtain the abnormal parameter Hy of the multi-modal environmental data;

[0080] Fire risk assessment: Input the fused multi-modal environmental information into the fire risk assessment model, predict the probability and potential location of fire occurrence, calculate the fire risk coefficient. The probability of fire occurrence, potential location, and fire risk coefficient constitute a fire risk report, and set safety protection measures based on the fire risk report; Automatically trigger intelligent responses (such as local isolation or fire extinguishing agent spraying).

[0081] For easy understanding, the definition method of the quantization threshold for the abnormal parameters of the temperature gradient and the temperature trend risk is as follows:

[0082] Based on historical normal operation data, calculate the 99% quantile of the temperature gradient mutation amplitude (the change rate of temperature difference between adjacent monitoring points) and the temperature trend risk slope (the temperature rise rate per unit time of the LSTM prediction curve) respectively as the initial anomaly threshold (that is, exceeding 99% of the normal fluctuation range is regarded as an anomaly), and construct a joint probability model of multi-modal environmental parameters (such as CO concentration, humidity extreme value) by combining with a Bayesian network: calculate the conditional probability of fire occurrence under each parameter combination by statistically analyzing the distribution of environmental parameters before a fire event, and set P≥0.75 as the high-risk threshold; further optimize the threshold through backtesting verification: using historical fire events as positive samples, adjust the quantile (such as 98%→99.5%) and the probability threshold (such as 0.75→0.8) to make the recall rate >90% and the false alarm rate <5%;

[0083] Dynamically weight key parameters (such as the weight of the temperature rise rate > CO concentration) through feature importance analysis (such as SHAP value), and solidify the lightweight Bayesian network to edge devices to achieve dynamic calibration of the threshold and hardware linkage execution.

[0084] The calculation formula for the temperature gradient anomaly parameter is: calculate the temperature difference between adjacent sensors divided by the distance between them to obtain the local temperature gradient; take the average of the sum of the squares of all local temperature gradients, and then take the square root of the result to obtain the temperature gradient anomaly parameter W y ;

[0085] The calculation method for the multi-modal environmental data anomaly parameter is: extract key features from the denoised multi-modal environmental data, such as vibration frequency, sound intensity, gas concentration value; set the normal range threshold for each key feature according to historical data and safety standards; calculate the anomaly degree of each key feature; sum the anomaly degrees of each key feature after weighting and perform normalization to obtain the multi-modal environmental data anomaly parameter.

[0086] In the embodiments of the present invention, it needs to be further explained that the response strategy is dynamically optimized to ensure both resource utilization efficiency and safety. The target monitoring battery is divided into several regions, and the maintenance priority is determined based on the fire risk coefficient. The acquisition method of the fire risk coefficient is as follows:

[0087] Obtain the fire probability P fire , which is the fire occurrence probability predicted by the LSTM model within the future time, and the value range is [0, 1];

[0088] Combine the temperature trend risk T risk , the multi-modal environmental data anomaly parameter H y and the fire probability, and obtain the fire risk coefficient C through the following formula risk =1 - exp[-k·(p fire ·(1 + T risk )·(1 + Hy ))], where k is a constant controlling the overall risk sensitivity, with a range of [1, 10].

[0089] Explanation: In the present invention, the LSTM model is used to predict the probability of a fire occurring in a future time period, and its data sources are historical fire event data and real-time multi-modal environmental data;

[0090] The historical data is extracted from the operation logs and accident reports of the energy storage system, and includes parameters such as temperature, vibration, sound, and gas concentration at the time of the fire;

[0091] The real-time data is collected through a sensor network, including multi-modal environmental information monitored by temperature sensors, vibration sensors, sound sensors, and gas concentration sensors;

[0092] The collected raw data is preprocessed through denoising, normalization, and feature extraction, and organized in the form of a time series;

[0093] The input for each time step is a vector containing multi-dimensional features such as temperature, temperature gradient, vibration frequency, sound intensity, and gas concentration. The input data is provided to the LSTM model in the form of a three-dimensional tensor ([batch size, sequence length, number of features]);

[0094] The LSTM model is trained based on historical data to learn the relationship between the time series and fire events, and predict the fire probability; during operation, the real-time data is input into the LSTM model after the same preprocessing, and the fire occurrence probability is output.

[0095] In a possible embodiment, the time step of the LSTM model is set to 1 minute, and the edge computing device processes the historical data of the most recent 30 minutes to form an input tensor; a 2-layer stacked LSTM structure is adopted, with each layer containing 128 hidden units, and the connection layer combines the sigmoid activation function to output the probability of a fire occurring.

[0096] In a possible embodiment, the fire risk assessment model is deployed on the edge device to achieve millisecond-level real-time warning; the cloud is responsible for the continuous training and updating of the model, uses large-scale historical data to improve the model accuracy, and adapts to the operating characteristics of different energy storage systems at the same time; according to the risk assessment results (fire risk coefficient), different fire protection strategies are automatically executed, such as:

[0097] Low risk: Locally strengthen cooling;

[0098] Medium risk: Isolate the fault area;

[0099] High risk: Trigger the spraying of fire extinguishing agent or emergency cooling.

[0100] In the embodiments of the present invention, it needs to be further explained that the method further includes an operation optimization management step, including: constructing a multi-objective optimization mathematical model based on the energy efficiency ratio, the full-life cycle benefit, and the battery health; using the multi-objective optimization mathematical model, combining the operation data of the energy storage system and the grid demand, and through constructing the multi-objective optimization mathematical model to quantify the operation characteristics, initializing the non-dominated sorting genetic algorithm population to generate a preliminary strategy, and performing evolutionary computation to generate a Pareto optimal solution set to achieve multi-objective collaborative optimization, optimizing the charge and discharge strategy and the parameters of the liquid cooling system, and generating a Pareto optimal solution set with economy, reliability, and environmental friendliness; introducing the Bayesian optimization technology to dynamically adjust the algorithm parameters to improve the optimization efficiency and the global search ability.

[0101] In the embodiments of the present invention, it needs to be further explained that, referring to Figure 2 the temperature control flow chart based on the operation strategy optimization, the operation optimization management step includes the following contents:

[0102] Step S31, constructing a multi-objective optimization mathematical model: extracting the grid demand and key parameters of the energy storage system from the operation data, the grid demand data at least includes the load and the frequency regulation demand, and the key parameters include the energy efficiency ratio, the full-life cycle benefit, and the battery health, and jointly establishing a multi-objective optimization mathematical model; outputting the multi-objective optimization mathematical model and the grid demand data;

[0103] Explanation: The energy efficiency ratio refers to the ratio of the heat removed by the liquid cooling system to the input energy; the full-life cycle benefit refers to the balance of the benefit and cost calculated based on exponential decay discounting; the battery health refers to the life assessment based on the capacity decay rate;

[0104] Step S32, initializing the non-dominated sorting genetic algorithm population: based on the multi-objective optimization mathematical model and the grid demand data, initializing the population using the non-dominated sorting genetic algorithm, and outputting the initial operation strategy set;

[0105] Explanation: randomly generating multiple groups of decision variable combinations through the non-dominated sorting genetic algorithm to construct the initial population, and each group of combinations corresponds to a potential operation strategy; the decision variables include: the charge and discharge power, the liquid cooling parameters (coolant flow rate, temperature setting value), and the grid demand;

[0106] And setting physical constraint conditions (such as power range, flow rate range) to ensure feasibility; based on the multi-objective optimization mathematical model, calculating the objective values of each population individual, including the energy efficiency ratio, the full-life cycle benefit, and the battery health, to form the initial operation strategy set;

[0107] Step S33, evolving to generate a Pareto optimal solution set: based on the initial operation strategy set, performing evolutionary computation using the non-dominated sorting genetic algorithm, and outputting the Pareto optimal solution set;

[0108] Explanation: Identify the Pareto front layer, where the first layer is the current optimal strategy; through the tournament selection mechanism, select the parent strategies from the population according to the crowding distance and domination rank; use simulated binary crossover and polynomial mutation to adjust the decision variables to generate offspring strategies and expand the diversity of the solution space; iteratively execute the sorting, selection, and mutation processes until the convergence condition is met (such as the number of iterations reaches the preset value or the front layer is stable) to generate the Pareto optimal solution set; each set of solutions includes the optimized charge and discharge power and liquid cooling parameters, corresponding to the target values of the energy efficiency ratio, the life cycle benefit, and the battery health.

[0109] Step S34, Bayesian optimization adjustment: Based on the Pareto optimal solution set, introduce Bayesian optimization technology to adjust the crossover probability and mutation index of the non-dominated sorting genetic algorithm to generate the final operation optimization control strategy to improve the performance of the solution set and the global optimization ability.

[0110] Explanation: Identify the hypervolume index (measuring diversity and convergence) of the Pareto optimal solution set, construct a Gaussian process model, and predict the parameter adjustment effect based on the historical combination of crossover probability and mutation index and its hypervolume index value; through the expected improvement function, determine the next set of optimal parameters from the prediction results, apply them to the non-dominated sorting genetic algorithm, and re-evolve and update the Pareto optimal solution set; iterate this adjustment process until the increase in the hypervolume index is less than the preset threshold, confirm the optimization convergence, and finally generate the Pareto optimal solution set as the operation optimization control strategy to guide the charge and discharge of the system and the configuration of liquid cooling parameters.

[0111] In a possible embodiment, when simultaneously optimizing the energy efficiency ratio (short-term index), the life cycle benefit (long-term index), and the battery health (reliability index), there may be contradictions in the multi-objective optimization mathematical model (for example, high energy efficiency may accelerate battery aging). The multi-objective conflict is balanced through the dynamic weight allocation and hierarchical constraint mechanism to solve the problem that the multi-objective optimization mathematical model may not converge or the actual effect deviates from the expectation. For the conflict problem of the energy efficiency ratio (short-term), the life cycle benefit (long-term), and the battery health (reliability), in the existing NSGA-II and Bayesian optimization frameworks, the dynamic weight allocation and hierarchical constraint mechanism are introduced to ensure the balance between the objectives and the convergence of the algorithm:

[0112] Dynamic weight allocation: Automatically adjust the weights of the energy efficiency ratio, benefit, and health in the objective function according to the power grid demand scenario (such as peak / valley load) and the battery aging state; for example, increase the weight of the energy efficiency ratio to 50% during peak load and increase the weight of the life cycle benefit to 20% during valley load; if the battery health deteriorates rapidly due to aging, automatically reduce the energy efficiency weight to give priority to ensuring the health and avoid overcharging and discharging to reduce the life.

[0113] Hierarchical constraint mechanism: Regarding the health status as a hard constraint (non-compromisable indicator), and the energy efficiency ratio and revenue as soft goals (optimizable indicators), to avoid excessive sacrifice of battery life; setting a minimum safety threshold for battery health (such as capacity attenuation not exceeding 20%), and all optimization strategies must meet this condition to ensure that the reliability bottom line cannot be breached; on the premise that the health status meets the standard, allowing flexible trade-off between energy efficiency and revenue, for example, achieving the balance between short-term energy efficiency and long-term revenue by adjusting cooling parameters or charge-discharge power;

[0114] Bayesian optimization extension: Incorporating weight parameters and constraint thresholds into the hyperparameter space of Bayesian optimization to achieve adaptive adjustment of global strategies; when generating Pareto optimal solutions, preferentially retaining the solution sets with qualified health status to ensure that the final strategies take into account both safety and economy.

[0115] Explanation: The dynamic weight responds to external scenario changes, the hierarchical constraint holds the safety bottom line, and Bayesian optimization globally coordinates the relationship among the three, forming a closed loop of "scenario perception → goal balance → strategy generation → parameter iteration", which not only avoids algorithm divergence but also fits the actual needs.

[0116] In the embodiments of the present invention, it needs to be further explained that the method further includes a fault location step:

[0117] If any one of the abnormal temperature gradient parameters, temperature trend risks, and abnormal multimodal environment data parameters exists, trigger the fault location step;

[0118] Using a time series anomaly detection algorithm to monitor operation data in real time and automatically identify abnormal behaviors; combining causal inference technology to analyze the causal relationship between anomalies and components (such as batteries and liquid cooling units) to quickly locate the root cause; constructing a fault knowledge graph to automatically classify abnormal patterns based on historical data and recommend solutions;

[0119] Explanation: Using an RNN model based on the attention mechanism, with the input being the time series data of temperature, voltage, and current, and the output being the anomaly score. When the anomaly score exceeds the preset value, it is marked as an abnormal behavior. An example formula for the anomaly score:

[0120]

[0121] Among them, are the predicted values of temperature, voltage, and current, σ is the standard deviation of historical data, and λ is the sensitivity factor.

[0122] Causal inference: Adopting causal inference technology, defining causal graph nodes (such as "liquid cooling pump failure" and "battery aging"), and analyzing the root cause through gradient backpropagation.

[0123] Knowledge Graph Update: Dynamically update the fault knowledge graph using graph neural networks. Rule extraction is based on the Apriori algorithm with a minimum support of 0.1 and a confidence threshold of 0.8.

[0124] Obtain the operating data of the energy storage system, apply a recurrent neural network based on the attention mechanism to analyze the temporal features, generate an abnormal behavior sequence to identify potential fault signals (such as abnormal temperature sudden increase or voltage sudden drop, etc.);

[0125] Obtain the abnormal behavior sequence, combine the structural causal model to infer the causal relationship between component faults and abnormal behavior sequences, generate a fault causal graph to locate the root cause of the fault; including:

[0126] Starting from the abnormal behavior sequence and historical operating data, use graph neural networks to construct a fault knowledge graph. The nodes in the graph represent system components (such as battery modules, liquid cooling units) and abnormal types (such as overheating, short circuit), and the edges represent causal relationships (for example, "battery aging → overheating"); use graph embedding technology to map the graph structure to a low-dimensional vector space (such as 128 dimensions), and train the edge weights (range 0 to 1) based on historical data to reflect the association strength between various faults; through iterative optimization of the graph embedding loss function, update the node vectors to generate an optimized fault knowledge graph, making similar abnormal patterns closer and different patterns more distinct;

[0127] Obtain the fault causal graph, combine graph neural networks and fuse historical data to generate a fault diagnosis report to classify abnormalities and recommend solutions.

[0128] Explanation: The initial fault knowledge graph is constructed based on historical fault cases. The nodes include fault types (such as overheating, short circuit) and components (such as battery modules, liquid cooling pumps);

[0129] The fault knowledge graph is updated through online learning: when a new fault occurs, extract features and match existing nodes. If there is no match, new nodes are added, and the edge weights are optimized through gradient descent; Example: "If the frequency of detecting 'abnormal liquid cooling pump flow' and 'local overheating' occurring simultaneously exceeds the threshold, then add a new edge 'liquid cooling pump failure → uneven cooling → temperature anomaly' with an initial weight of 0.9".

[0130] In a possible embodiment, an abnormal behavior sequence is identified based on an abnormal behavior detection threshold function, and the calculation formula of the abnormal behavior detection threshold function is:

[0131]

[0132] where, x t is the current data value, x pred represents the predicted value, x ref is the reference value, and λ represents the sensitivity factor.

[0133] Summary: The embodiments of the present invention combine battery electrochemistry models, sensor networks, long short-term memory networks, fire risk assessment, and operation optimization management technologies to provide a comprehensive safety protection and energy efficiency optimization strategy. By real-time monitoring data such as battery temperature, current, and voltage, predicting the temperature change trend, and dynamically adjusting the control strategy of the liquid cooling system, it ensures that the energy storage system operates efficiently while avoiding safety risks such as overheating. In addition, the fire risk assessment model combines multi-modal environmental data such as temperature, vibration, and gas concentration to trigger intelligent response measures in a timely manner, improving the safety of the system. The present invention effectively solves the deficiencies of traditional energy storage systems in energy efficiency management and safety protection, and improves the operation efficiency and safety guarantee ability of the system.

[0134] Embodiment 2. The difference between the embodiment of the present invention and Embodiment 1 is that it further includes the step of optimizing the layout of the liquid cooling pipeline.

[0135] If any one of the abnormal temperature gradient parameters, temperature trend risks, and multi-modal environmental data abnormal parameters still exists in the scenario of the operation optimization control strategy, trigger the step of optimizing the layout of the liquid cooling pipeline.

[0136] In the embodiments of the present invention, it needs to be further explained that the process of optimizing the layout of the liquid cooling pipeline includes the following steps:

[0137] Obtain the battery temperature change trend curve and the real-time temperature distribution matrix, input the heat transfer parameters into the heat flow simulation model, and use reinforcement learning to iteratively adjust the pipe diameter, branches, and flow distribution. Based on the cooling uniformity, liquid cooling energy efficiency, and the highest temperature, set the reward target for pipeline optimization; generate the optimized pipeline design parameters to ensure both cooling uniformity and energy efficiency.

[0138] Obtain the pipeline design parameters, adjust the layout of the liquid cooling pipeline, and verify the battery temperature change trend curve and the real-time temperature distribution matrix to ensure the reliability of pipeline optimization.

[0139] Explanation: The heat flow simulation model is a digital twin model used to simulate the heat transfer and distribution characteristics of battery modules in a liquid-cooled industrial and commercial energy storage system. By dividing the battery module into finite element meshes, calculating the temperature change of each unit based on the heat conduction equation, and simulating the impact of liquid cooling pipeline adjustments (such as pipe diameter and flow rate) on the heat flow; input the distribution map of uneven cooling areas and heat transfer parameters, and output the optimized pipeline design parameters and the corresponding predicted temperature distribution.

[0140] In the embodiments of the present invention, it needs to be further explained that the calculation method of the reward target is as follows:

[0141] Obtain the cooling uniformity, the higher the value, the better the uniformity; obtain the cooling efficiency Q remored and the energy consumption E coolingBy comparing, the liquid cooling energy efficiency is calculated; the reward target is calculated through the following formula:

[0142]

[0143] wherein, T i represents the temperature at a certain space-time position, w1, w2, and w3 respectively represent the weight coefficients of each item, T max is the highest temperature, T safe is the safety threshold, and T ref is the reference temperature.

[0144] Summary: On the basis of Embodiment 1, steps for optimizing the layout of the liquid cooling pipeline are added, further improving the thermal management efficiency of the liquid cooling system; specifically, when abnormal temperature gradient parameters, temperature trend risks or abnormal multimodal environment data parameters are detected, the liquid cooling pipeline optimization is automatically triggered, and the diameter, branches and flow distribution are iteratively adjusted by using a reinforcement learning algorithm to ensure the balance between cooling uniformity and liquid cooling energy efficiency. Through the heat flow simulation model, the heat transfer and distribution characteristics in the liquid cooling system are simulated. The optimized pipeline design not only improves the cooling uniformity of the liquid cooling system, but also reduces the energy consumption, effectively solving the problems of thermal non-uniformity and low energy efficiency caused by the unreasonable layout of the liquid cooling pipeline, and improving the safety and operation efficiency of the energy storage system.

[0145] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A collaborative method for energy management and safety protection of a liquid-cooled industrial and commercial energy storage system, characterized in that, Including: Simulate the electrochemical reaction and thermal behavior of the battery to generate a preset thermal distribution map and heat transfer parameters; Collect the temperature, voltage, current data of the battery and multi-modal environment data in real time through the sensor network. Filter the noise by the edge computing device and calculate the temperature gradient, and output the real-time temperature distribution matrix and temperature gradient data; Build a temperature trend prediction model using a long short-term memory network. Input the preset thermal distribution map, heat transfer parameters, real-time temperature distribution matrix and electrical parameters, and output the battery temperature change trend curve. Set the liquid cooling control strategy based on the battery temperature change trend curve; Dynamically adjust the coolant flow rate and temperature of the liquid cooling system; Fuse the temperature gradient anomaly parameter, temperature trend risk and multi-modal environment data anomaly parameter, start the fire risk assessment model, calculate the fire risk coefficient, generate a fire risk report and trigger safety protection measures; Construct a multi-objective optimization mathematical model based on the energy efficiency ratio, full life cycle benefit and battery health. Combine the energy storage operation data with the grid demand, and use the non-dominated sorting genetic algorithm to generate the Pareto optimal solution set. Optimize the charge and discharge strategy and liquid cooling parameters, and introduce the Bayesian optimization technology to dynamically adjust the algorithm parameters; If the temperature remains abnormal, use reinforcement learning to optimize the liquid cooling pipeline layout, iteratively adjust the pipe diameter and flow distribution to ensure cooling uniformity and energy efficiency.

2. The energy management and safety protection collaborative method of a liquid-cooled industrial and commercial energy storage system according to claim 1, characterized in that, Based on the battery temperature change trend curve, adaptively adjust the acquisition frequency of battery data; The long short-term memory network converts the spatial data into a time series format, including dividing the preset thermal distribution map into regions and extracting the average temperature of each region to form a time series input.

3. The energy management and safety protection collaborative method for a liquid-cooled industrial and commercial energy storage system according to claim 1, characterized in that, The triggering conditions of the fire risk assessment model are at least one of the following: The temperature gradient anomaly parameter exceeds the preset threshold; The temperature trend risk result exceeds the dynamically adjusted threshold; the temperature trend risk T risk refers to the risk accumulation from the current moment t to a future time period t + Δt, and is calculated by the following formula: Among them, represents the rate of change of temperature with time; represents the acceleration of temperature change, β1 is a tuning parameter used to amplify or attenuate the influence of acceleration; σ pred represents the temperature prediction confidence interval, reflecting the credibility of future trends; β2 represents the uncertainty weight, used to adjust the contribution of prediction errors to risk; The multi-modal environment data anomaly parameter exceeds the threshold through weighted fusion of the anomaly degrees of vibration, sound and gas concentration.

4. The energy management and safety protection coordination method of a liquid-cooled industrial and commercial energy storage system according to claim 3, wherein, The method includes fire risk management steps, including: Environmental data collection: The sensor network also collects the vibration, sound and gas concentration data of the battery, and fuses them to obtain multi-modal environment data; Abnormality discrimination: Trigger the fire risk assessment model based on the temperature gradient anomaly parameter, temperature trend risk and multi-modal environment data anomaly parameter. The temperature trend risk refers to the potential overheating risk of the battery temperature in a future period based on the predicted temperature change trend; The multi-modal environment data anomaly parameter refers to the degree to which vibration, sound and gas concentration exceed the preset range; Fire risk assessment: Input the fused multi-modal environment information into the fire risk assessment model, predict the probability and potential location of fire occurrence, calculate the fire risk coefficient. The probability of fire occurrence, potential location and fire risk coefficient constitute the fire risk report, and set safety protection measures based on the fire risk report.

5. The energy management and safety protection coordination method of a liquid-cooled industrial and commercial energy storage system according to claim 4, characterized in that, Divide the target monitored battery into several regions, and determine the maintenance priority based on the fire risk coefficient. The acquisition method of the fire risk coefficient is: Obtain the fire probability P fire , which is the fire occurrence probability within the future time predicted by the LSTM model, and its value range is [0, 1]; Combined temperature trend risk T risk , abnormal parameter H of multi-modal environmental data y and fire probability, the fire risk coefficient C is obtained through the following formula risk = 1 - exp[-k·(P fire ·(1 + T risk )·(1 + H y ))], where k represents a constant controlling the overall risk sensitivity, in the range [1, 10].

6. The energy management and safety protection collaborative method for a liquid-cooled industrial and commercial energy storage system according to claim 1, characterized in that It also includes an operation optimization management step, which includes: constructing a multi-objective optimization mathematical model based on the energy efficiency ratio, full life cycle revenue, and battery health; using the multi-objective optimization mathematical model, combining the operation data of the energy storage system and the grid demand, and through constructing the multi-objective optimization mathematical model to quantify the operation characteristics, initializing the non-dominated sorting genetic algorithm population to generate a preliminary strategy, and performing evolutionary calculation to generate a Pareto optimal solution set to achieve multi-objective collaborative optimization, optimizing the charge and discharge strategy and the parameters of the liquid cooling system, and generating a Pareto optimal solution set; introducing the Bayesian optimization technology to dynamically adjust the algorithm parameters to improve the optimization efficiency and global search ability.

7. The energy management and safety protection collaborative method of a liquid-cooled industrial and commercial energy storage system according to claim 6, wherein The operation optimization management step includes the following contents: Step S31, constructing a multi-objective optimization mathematical model: extracting the grid demand and key parameters of the energy storage system from the operation data, the grid demand data at least includes the load and frequency regulation demand, and the key parameters include the energy efficiency ratio, full life cycle revenue, and battery health, and jointly establishing a multi-objective optimization mathematical model; outputting the multi-objective optimization mathematical model and the grid demand data; Step S32, initializing the non-dominated sorting genetic algorithm population: based on the multi-objective optimization mathematical model and the grid demand data, initializing the population using the non-dominated sorting genetic algorithm, and outputting the initial operation strategy set; Step S33, evolving to generate a Pareto optimal solution set: based on the initial operation strategy set, performing evolutionary calculation using the non-dominated sorting genetic algorithm, and outputting a Pareto optimal solution set; Step S34, Bayesian optimization adjustment: based on the Pareto optimal solution set, introducing the Bayesian optimization technology to adjust the crossover probability and mutation index of the non-dominated sorting genetic algorithm, and generating the final operation optimization control strategy.

8. A collaborative method for energy management and safety protection of a liquid-cooled industrial and commercial energy storage system according to claim 7, characterized in that, It also includes the step of optimizing the layout of the liquid cooling pipeline: If any one of the abnormal temperature gradient parameters, temperature trend risk, and multi-modal environment data abnormal parameters still exists in the operation optimization control strategy scenario, the step of optimizing the layout of the liquid cooling pipeline is triggered.

9. The energy management and safety protection collaborative method of a liquid-cooled industrial and commercial energy storage system according to claim 8, characterized in that, The process of optimizing the layout of the liquid cooling pipeline includes the following steps: Inputting the heat transfer parameters and the temperature distribution matrix into the heat flow simulation model; Setting the reward target based on the cooling uniformity, energy efficiency, and the highest temperature; Iteratively adjusting the pipe diameter and branches until the cooling uniformity requirement is met.

10. A method for coordinating energy management and safety protection of a liquid-cooled industrial and commercial energy storage system according to claim 9, characterized in that, The calculation method of the reward target is: Obtain the cooling uniformity, where the higher the value, the better the uniformity; obtain the cooling efficiency Q remored and the energy consumption E cooling ratio, and calculate the liquid cooling energy efficiency; obtain the reward target through the following formula; Among them, T i represents the temperature at a certain space-time position, w1, w2, and w3 respectively represent the weight coefficients of each item, T max is the highest temperature, T safe is the safety threshold, T ref is the reference temperature.

Images