Safety intelligent management system of building construction site mobile energy storage power supply

Abstract

The invention relates to the technical field of power supply safety intelligent management, and discloses a safety intelligent management system of a building construction site mobile energy storage power supply, which comprises a multi-physical field sensing layer, an edge computing control layer, an active safety execution layer and a cloud intelligent platform layer. Data are collected in real time through a temperature / potential / stress sensor array, three-dimensional physical field distribution is reconstructed by adopting a Kriging algorithm, and a thermal runaway risk index is calculated. Hardware-level redundancy is achieved through the heterogeneous design of the main control MCU and the monitoring MCU, and the safety mode is switched when the double-computer deviation exceeds the limit. The phase change cooling unit is combined with the magnetofluid circuit breaking mechanism to quickly respond to thermal runaway, and the self-destruction pressure relief unit directionally releases pressure. And the cloud digital twin model synchronously updates the three-dimensional thermal field, and the LSTM network predicts life attenuation. A super capacitor is configured for emergency power supply, and key parts are maintained to operate within 20ms after a main power supply is interrupted. According to the method, the problems of high single-point failure rate, unstable standby power and inaccurate thermal field monitoring in a traditional scheme are solved, and the early thermal runaway recognition rate and the system fault-tolerant capability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power safety intelligent management, and particularly to a safety intelligent management system for a mobile energy storage power supply at a building construction site. Background Art

[0002] Currently, the battery safety management system generally adopts a centralized control architecture, and the main control unit simultaneously processes data acquisition, risk assessment, and instruction issuance. Under this architecture, once an arithmetic error or communication failure occurs in the main control chip, the error instruction will be directly transmitted to the execution layer, lacking a real-time error correction mechanism at the hardware level. Most existing fault tolerance solutions set heartbeat detection at the software layer, but cannot intercept transient abnormal signals, resulting in a risk of mis-triggering key operations such as cooling and power-off.

[0003] In the design of emergency power supply, the mainstream solution relies on a lithium battery pack as a backup power source. The lithium battery has obvious characteristics of a sharp drop in capacity in a low-temperature environment, and long-term floating charge state will accelerate battery aging. When the main power supply suddenly interrupts, the output voltage of the lithium battery fluctuates greatly, making it difficult to maintain the stable operation of the sensor network and the control circuit, and may cause loss of monitoring data or delay in protection actions.

[0004] Most of the thermal runaway warning methods are based on fixed temperature thresholds or single-parameter trend analysis, and fail to effectively integrate multi-dimensional data such as temperature field and stress field. In actual operation, the thermo-electro-mechanical coupling effect inside the battery results in a complex heat diffusion path. Only monitoring the local temperature peak is likely to miss the hidden risk caused by interlayer heat accumulation. Traditional algorithms are insensitive to the change of three-dimensional temperature gradient and are difficult to capture the abnormal characteristics at the initial stage of the formation of micro-area hot spots in a timely manner.

[0005] The temperature monitoring network usually adopts a sparse layout to reduce costs. The too large distance between adjacent sensors limits the accuracy of heat field reconstruction. The discrete point data is difficult to accurately restore the overall heat distribution on the battery surface. Especially when thermal runaway occurs in the blind area covered by the sensor, the system cannot quickly locate the fault point, delaying the startup time of the protection mechanism. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a safety intelligent management system for a mobile energy storage power supply at a building construction site, which solves the problem of safety protection failure in battery safety management caused by single-point failure, unstable emergency power supply, lagging thermal runaway warning, and incomplete thermal field monitoring.

[0007] To achieve the above object, the present invention is realized through the following technical solutions: A safety intelligent management system for a mobile energy storage power supply at a building construction site, comprising:

[0008] A multi-physical field perception layer for real-time collecting data of the temperature field, electric potential field, and mechanical stress field of the battery pack;

[0009] The edge computing control layer, which is communicatively connected to the multi-physical field perception layer, is used to reconstruct the physical field distribution and calculate the thermal runaway risk index;

[0010] The active safety execution layer receives the instructions from the edge computing control layer and executes cooling control and circuit protection operations;

[0011] The cloud intelligent platform layer communicates bidirectionally with the edge computing control layer to achieve remote status monitoring and policy optimization.

[0012] Preferably, the multi-physical field perception layer includes:

[0013] The temperature sensing unit is composed of a flexible sensor array. At least 15 temperature detection points are arranged on the surface of each battery cell in an icosahedron topology, with a detection accuracy of ±0.5°C, and is used to generate three-dimensional temperature field distribution data;

[0014] The electric potential detection unit includes 3 to 8 thin-film voltage sensors, which are distributed along the diagonal or in a grid pattern of the battery cell, and is used to measure the electric potential gradient and identify local overvoltage areas;

[0015] The deformation sensing unit uses 5 to 10 MEMS micro-strain sensors, which are embedded in the battery housing in a cross layout, and is used to detect stress anomalies caused by mechanical extrusion or deformation, with a sensitivity of 30 to 100 μm / m.

[0016] Preferably, the temperature sensing unit includes a data preprocessing module, which includes:

[0017] A 2^4-bit high-precision analog-to-digital converter, which is used to convert analog signals into digital signals, and the sampling frequency is not less than 1 kHz;

[0018] The digital filtering unit is configured to suppress noise above 50 Hz through a Butterworth low-pass filter, and the passband attenuation is less than 3 dB.

[0019] Preferably, the edge computing control layer includes:

[0020] The field reconstruction unit converts discrete sensor data into a continuous temperature field based on the Kriging spatial interpolation algorithm. The weight coefficient is calculated by the Gaussian function of the sensor spacing, and the correlation length is set to 10 mm to 20 mm, and is used to accurately restore the thermal field distribution;

[0021] The risk prediction unit calculates the thermal runaway risk index through integral operation. The index includes an integral term of the square of the temperature change rate and an integral term of the square of the temperature gradient. The weighting coefficients of the two terms are 0.2 to 0.5 and 0.8 to 1.5 respectively, and are used to quantify the thermal runaway risk level;

[0022] The decision-making control unit triggers the cooling power adjustment instruction and the emergency power-off signal according to the risk index grading.

[0023] Preferably, the field reconstruction unit includes:

[0024] An FPGA parallel computing architecture for accelerating the update of the three-dimensional temperature field, with a computing cycle not exceeding 200 ms;

[0025] An adaptive mesh generation module that automatically increases the local mesh density to 1 mm accuracy when the detected temperature gradient exceeds 3 °C / cm to 8 °C / cm, for capturing details of heat propagation.

[0026] Preferably, the active safety execution layer includes:

[0027] A phase change cooling unit composed of a microchannel ejector and a phase change material storage tank, with a flow control range of 0 to 10 L / min and a response time not exceeding 50 ms, for dynamically adjusting the cooling intensity according to the risk level;

[0028] A magnetohydrodynamic circuit breaker unit with a liquid metal contact design that cuts off the charge and discharge circuit above 150 A within 20 ms after receiving a secondary alarm signal, for preventing the spread of thermal runaway;

[0029] A self-destruction pressure relief unit equipped with a shape memory alloy valve that directionally opens the pressure relief channel when the internal pressure exceeds 2 kPa to 5 kPa, for quickly releasing the internal gas pressure.

[0030] Preferably, the control strategy of the phase change cooling unit satisfies:

[0031] The cooling power consumption has a quadratic function relationship with the flow rate, with a proportionality coefficient of 0.1 to 0.3, for optimizing energy consumption;

[0032] The temperature deviation penalty term is proportional to the square of the difference between the measured temperature and the safety threshold, with a penalty coefficient of 5 to 10, for balancing the cooling efficiency and temperature stability.

[0033] Preferably, the cloud intelligent platform layer includes:

[0034] A digital twin unit for constructing a three-dimensional simulation model based on the physical field data uploaded by the edge layer, with a data update cycle not exceeding 10 seconds and a surface convective heat transfer coefficient of 8 W / (m 2 ·K) to 15 W / (m 2 ·K);

[0035] A remaining useful life prediction unit that analyzes the historical thermal risk index sequence and capacity attenuation rate through an LSTM neural network, with a prediction error of the remaining useful life not exceeding 5%;

[0036] A remote locking unit that adopts a three-factor authentication mechanism of device fingerprint, dynamic token, and geographical location, and sends an encrypted locking instruction after successful verification.

[0037] Preferably, the implementation manner of the digital twin unit is as follows:

[0038] A finite element solver for solving the thermoelectric coupling equation, and the boundary conditions include electrode insulation and surface convective heat transfer;

[0039] A real-time rendering module that maps the temperature field data into a pseudo-color three-dimensional model with a color scale resolution of 0.3 °C to 1.0 °C for visual monitoring.

[0040] Preferably, it further includes:

[0041] A dual MCU redundant control module, where the main control MCU performs real-time field reconstruction and risk calculation, and the monitoring MCU independently verifies data consistency. When the output deviation exceeds 3% to 8%, it switches to the safe mode to prevent single-point failure;

[0042] An emergency power supply unit configured with a supercapacitor energy storage device to maintain the operation of sensors, communication modules, and key control circuits for at least 15 minutes after the main power supply is interrupted, for ensuring continuous monitoring in the emergency state of the system.

[0043] The present invention provides a safety intelligent management system for a mobile energy storage power supply at a construction site, having the following beneficial effects:

[0044] 1. Through the heterogeneous architecture design of the main control MCU and the monitoring MCU, combined with the dynamic threshold arbitration mechanism, the present invention forms a hardware-level redundant protection. The main control MCU performs real-time field reconstruction and risk calculation, and the monitoring MCU independently verifies data consistency. When the output deviation between the two machines exceeds 3% - 8%, it is forced to switch to the preset safe mode through a hardware watchdog. Traditional solutions rely on a single controller and software verification, with risks of missed detection and response delay. The present invention can still maintain the basic functions of cooling, power-off, and pressure relief in case of single-point failure, achieving full-link fault tolerance from data acquisition to execution.

[0045] 2. The present invention adopts a carbon-based supercapacitor module and dynamic PID voltage regulation technology to switch to emergency power supply within 20 ms after the main power supply is interrupted, maintain the continuous operation of sensors, communication modules, and control circuits, suppress the 48V bus voltage fluctuation (≤5%) through a bidirectional Buck-Boost circuit, and be compatible with photovoltaic charging and lithium battery expansion. Traditional lithium battery backup solutions have defects such as low-temperature capacity attenuation and short cycle life. The present invention utilizes the fast charge and discharge characteristics of supercapacitors, combined with a dynamic load matching strategy, to ensure stable power support for more than 15 minutes in extreme environments.

[0046] 3. Based on the LSTM neural network and attention mechanism, the present invention integrates the temperature field, stress field, and historical capacity attenuation data to construct a thermal runaway risk prediction model. By using a real-time field reconstruction algorithm, the three-dimensional temperature gradient and electric potential distribution are obtained. Combining with the stress sensor data, the abnormal evolution law under the coupling action of multiple physical fields is identified. The traditional solution relies on a single temperature threshold or empirical formula and cannot capture the complex thermal-electro-mechanical coupling effect. The present invention realizes accurate early warning at the 15-minute level, achieving the purpose of reducing the false alarm rate.

[0047] 4. The present invention deploys more than 15 micro-sensor nodes on the surface of the battery cell. The anisotropic Kriging algorithm is used to reconstruct the three-dimensional temperature field, improving the spatial resolution. The heat diffusion path is monitored in real time through the gradient tracking algorithm. When the local temperature gradient exceeds 5 °C / cm, a three-level response mechanism is triggered. Due to insufficient sensor density, the traditional discrete point temperature measurement solution is difficult to capture the interlayer heat accumulation and small hot spots. The present invention improves the early thermal runaway recognition rate through three-dimensional field reconstruction and abnormal area positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is the framework diagram of the system of the present invention;

[0049] Figure 2 is the framework diagram of the multi-physical field perception layer of the present invention;

[0050] Figure 3 is the framework diagram of the edge computing control layer of the present invention;

[0051] Figure 4 is the framework diagram of the active safety execution layer of the present invention;

[0052] Figure 5 is the framework diagram of the cloud intelligent platform layer of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0054] Please refer to the attached Figure 1 , the embodiment of the present invention provides a safety intelligent management system for a mobile energy storage power supply at a construction site, including:

[0055] Please refer to the attached Figure 2 , a multi-physical field perception layer for real-time collection of temperature field, electric potential field, and mechanical stress field data of the battery pack;

[0056] In the system of the present invention, the multi - physical - field perception layer constitutes the basic unit for data acquisition. Through the collaborative layout and signal - processing mechanism of the heterogeneous sensor network, it realizes the real - time monitoring of the temperature field, electric - potential field, and mechanical - stress field of the battery pack. This layer is connected to the edge - computing control layer through a high - speed digital interface. The original sensing data is pre - processed and then transmitted to the upper layer for physical - field reconstruction and risk analysis, forming a complete closed - loop from signal acquisition to decision execution.

[0057] In this embodiment, the temperature - sensing unit is manufactured using flexible printed - circuit technology. The sensor array is arranged on the surface of the battery cell in an icosahedron topological structure. Generally, no less than 15 temperature - detection points are deployed on the surface of each battery cell, and the distance between adjacent detection points is controlled within the range of 10 - 20 mm to achieve high - resolution capture of the three - dimensional temperature gradient. As an option, the arrangement density of the detection points can be dynamically adjusted according to the size of the battery cell. For example, in a energy - storage unit with a capacity greater than 200 Ah, the number of detection points can be increased to more than 20.

[0058] Specifically, the process of generating temperature - field data follows the following technical route:

[0059] The analog signals output by each detection point are digitized by a 24 - bit high - precision analog - to - digital converter, and the sampling frequency is not less than 1 kHz. The integral non - linear error of the analog - to - digital converter is less than ±2 LSB, and when the input range is 0 - 5 V, the quantization error is lower than 0.1 mV.

[0060] The digital - filtering unit uses a fourth - order Butterworth low - pass filter to suppress high - frequency noise, and its transfer function is:

[0061]

[0062] where f c is the cut - off frequency, set to 50 Hz, used to filter out power - frequency interference and high - frequency noise, n is the filter order, taking a value of 4 to ensure that the stop - band attenuation rate is greater than 24 dB / octave, and s is the Laplace variable.

[0063] The pre - processed discrete temperature data is transmitted to the edge - computing layer through the SPI interface, and the data - transmission rate is not less than 10 Mbps, and the bit - error rate is controlled within 1×10 -9 below.

[0064] The temperature - field reconstruction uses an improved Kriging interpolation method, and its mathematical expression is:

[0065]

[0066] where, is the estimated temperature value at the target point (x, y, z), in degrees Celsius (°C), N is the total number of temperature detection points, that is, the number of sensors deployed on the surface of a single battery cell, and λ k is the weight coefficient, reflecting its contribution to the temperature estimation of the target point, T k is the preprocessed temperature value measured at the k-th temperature detection point, in degrees Celsius (°C);

[0067] The weight coefficient λ k is calculated by the following formula:

[0068]

[0069] where d k is the Euclidean distance between the target point and the k-th sensor, in millimeters (mm);

[0070] l T is the spatial correlation length of the temperature field, with a value of 15 mm, determined by the thermal diffusivity α of the battery cell material, and the calculation formula is:

[0071]

[0072] where α = 1.2×10 -5 m 2 / s (for lithium iron phosphate material), Δt = 100 ms (edge computing cycle).

[0073] In some embodiments, the sensor material is a platinum thin film resistor on a polyimide substrate, with a temperature sensitivity coefficient of 3850 ppm / °C and a non-linear error of less than ±0.15% in the range of -20°C to 120°C. The signal conditioning circuit integrates a temperature compensation module to eliminate the influence of environmental temperature drift through second-order polynomial fitting, and the compensation formula is:

[0074] T corrected = T raw + a(T amb - T ref ) + b(T amb - T ref ) 2 ;

[0075] where T amb is the environmental temperature, measured by an independent sensor, T ref is the reference temperature (25°C), and a and b are both compensation coefficients, determined to be a = 0.12 and b = 0.005 through calibration experiments.

[0076] In this embodiment, the potential detection unit is composed of 3 to 8 thin-film voltage sensors, which are distributed along the diagonal or in a grid pattern of the battery cell. Generally, the sensor spacing is not less than 1 / 3 of the side length of the battery cell to avoid electric field coupling interference. For a standard-sized battery cell (150 mm in length × 100 mm in width), the typical spacing is 30 - 50 mm.

[0077] Specifically, the potential gradient is calculated using the central difference method:

[0078]

[0079] where is the magnitude of the potential gradient vector at the target point (i, j), Δx and Δy are the adjacent sensor spacings, with the unit of millimeter (mm); when the detected local gradient value exceeds 5 mV / mm, an overvoltage warning is triggered. V i+1,j , V i-1,j are the measured values of two voltage sensors adjacent to the target point along the x-axis direction, with the unit of millivolt (mV), and V i,j+1 , V i,j-1 are the measured values of two voltage sensors adjacent to the target point along the y-axis direction, with the unit of millivolt (mV).

[0080] In a possible implementation, the sensor uses an indium tin oxide (ITO) transparent electrode with a sheet resistance less than 10 Ω / sq and a light transmittance greater than 85%, to avoid blocking the visual inspection of the internal structure. An instrumentation amplifier with a common-mode rejection ratio greater than 120 dB is added to the voltage sampling circuit to effectively suppress common-mode interference. The sampling circuit is designed to satisfy:

[0081] V out = G(V + - V - ) + V offset ;

[0082] where G is the gain, set to 100 times, and V offset is the offset voltage, which is controlled within ±50 μV by an automatic zeroing circuit.

[0083] In this embodiment, the deformation sensing unit is composed of 5 to 10 MEMS piezoresistive strain sensors, which are embedded inside the battery case in a cross layout. The sensitivity range is set to 30 - 100 μm / m to cover typical mechanical impact loads on a construction site (such as equipment collision and overloading).

[0084] Specifically, the stress anomaly detection is based on the following criterion:

[0085] σ max ≥ σ crit = min(0.5σ yield , 1.2σ oper );

[0086] Among them, σ max is the maximum equivalent stress monitored in real time, with the unit of megapascal, and σ yield is the yield strength of the housing material. For example, for an aluminum alloy housing, σ yield = 200 MPa, and σ oper is the allowable design stress, taking σ oper = 150 MPa, and σ crit is the critical stress threshold, with the unit of megapascal.

[0087] In some embodiments, the sensor package adopts a silicone gel filling structure, with an IP68 protection level, suitable for construction site dust and humid environments. The data fusion module jointly analyzes the strain data and the accelerometer signal to distinguish static loads and dynamic impact events. The accelerometer range is set to ±50 g, and the sampling rate is 1 kHz, meeting the requirements for impact event detection.

[0088] In this embodiment, the analog-to-digital converter adopts a Σ-Δ architecture, with an integral non-linearity error less than ±2 LSB and an effective number of bits not less than 21 bits at a sampling rate of 1 kHz. The implementation scheme of the digital filter unit includes:

[0089] The transfer function of the Butterworth filter:

[0090]

[0091] Among them, f c = 50 Hz, and n = 4.

[0092] In some embodiments, the preprocessing module embeds the FFT frequency domain analysis function to dynamically identify 50 Hz power frequency interference and its harmonic components, and generate adaptive notch filter coefficients. The data verification unit adds a CRC-16 cyclic redundancy check code, and the generating polynomial is:

[0093] G(x) = x 16 + x 15 + x 2 + 1;

[0094] The check code length is 16 bits, and the error rate is controlled below 1×10 -9 below.

[0095] The technical parameters of each sub-module and the subsequent processing flow form a strict mapping:

[0096] The temperature detection point density (≥15 per cell) matches the correlation length (15 mm) of the field reconstruction algorithm, ensuring that the spatial resolution meets the accuracy requirement of Δt < 0.5 °C;

[0097] The layout spacing of the voltage sensor is directly related to the calculation accuracy of the gradient. When Δx = 20 mm, the gradient detection resolution reaches 0.25 mV / mm;

[0098] The sensitivity of the strain sensor (30 - 100 μm / m) covers the elastic deformation stage of the shell deformation, avoiding irreversible damage caused by plastic deformation.

[0099] Please refer to the appendix Figure 3 , the edge computing control layer, is communicatively connected to the multi-physical field perception layer for reconstructing the physical field distribution and calculating the thermal runaway risk index;

[0100] In this embodiment, the field reconstruction unit converts the discrete temperature detection point data into a continuous three-dimensional temperature field distribution based on an improved Kriging spatial interpolation algorithm. Generally, the sensor data is transmitted to the edge computing node through the SPI interface at a rate not lower than 10 Mbps, and the data packet contains the position coordinates and temperature values.

[0101] The mathematical expression for temperature field reconstruction is:

[0102]

[0103] Among them, the weight coefficient λ k is calculated by the Gaussian function of the sensor spacing:

[0104]

[0105] Among them, d k is the Euclidean distance between the target point and the k-th sensor, with the unit of millimeter (mm);

[0106] l T is the spatial correlation length of the temperature field, with a value range of 10 mm to 20 mm, and the preferred value is 15 mm (determined jointly by the thermal diffusivity α = 1.2×10 -5 m 2 / s and the calculation period Δt = 100 ms, satisfying

[0107] In a possible implementation manner, the field reconstruction unit deploys a parallel computing architecture using an FPGA chip. 64 parallel multipliers are configured inside the FPGA, and 16 groups of weight coefficient calculations are completed in each clock cycle. The total process time for updating the three-dimensional temperature field does not exceed 200 ms, meeting the real-time requirement.

[0108] In some embodiments, the adaptive mesh generation module dynamically adjusts the mesh density according to the temperature gradient. When the temperature gradient in a local area is detected to exceed 3 °C / cm to 8 °C / cm, the mesh size in this area is switched from the default 5 mm to 1 mm. The gradient threshold is set based on:

[0109]

[0110] where ΔT max = 5 °C is the maximum allowable temperature difference between adjacent battery cells, and δx = 1.5 cm is the battery cell spacing.

[0111] In some embodiments, the field reconstruction algorithm introduces anisotropic correction and is applicable to non-uniform heat conduction scenarios:

[0112]

[0113] where is the correlation length in the x-axis direction, is the correlation length in the y-axis direction, is the correlation length in the z-axis direction (adapting to the heat conduction characteristics of the battery cell stacking structure), d k,x , d k,y , d k,z is the distance component of the k-th sensor and the target point in the x, y, and z-axis directions, with the unit of millimeters.

[0114] In this embodiment, the risk prediction unit quantifies the thermal runaway risk index through integral operation, and its mathematical expression is:

[0115]

[0116] where w1 is the weighting coefficient of the temperature change rate term, with a value range of 0.2 to 0.5, and the preferred value is 0.3; w2 is the weighting coefficient of the temperature gradient term, with a value range of 0.8 to 1.5, and the preferred value is 1.2. is the change rate of temperature with time, with the unit of °C / s, is the magnitude of the temperature spatial gradient, with the unit of °C / cm.

[0117] Specifically, the weighting coefficients are dynamically adjusted according to the state of charge (SOC) of the battery:

[0118] When SOC > 80%, w1 is increased to 0.4 to enhance the sensitivity to the temperature rate;

[0119] When SOC < 20%, w2 is decreased to 1.0 to reduce the weight ratio of the gradient term.

[0120] In some embodiments, the integral operation is discretized using the trapezoidal method:

[0121]

[0122] Wherein:

[0123]

[0124] Δt is the sampling interval;

[0125] Δx is the default grid size;

[0126] f(t n-1 ) and f(t n ) are the instantaneous risk contribution function values at times t n-1 and t n .

[0127] In some other embodiments, a stress coupling term is added to the risk prediction model:

[0128]

[0129] Wherein, w3 = 0.05, the force term weight coefficient is the output of the deformation sensing unit, the maximum equivalent stress, unit MPa.

[0130] In this embodiment, the decision control unit triggers different operations according to the classification threshold of the risk index Γ(t):

[0131] Level 1 alarm (Γ(t) ≥ 1500): Activate the phase change cooling unit, and the target flow rate is set to Q = 5 L / min;

[0132] Level 2 alarm (Γ(t) ≥ 2500): Cut off the magneto - fluid circuit breaker and start the pressure relief valve.

[0133] As an option, the flow control strategy adopts the PID algorithm:

[0134]

[0135] Wherein, e(t) = T meas (t) - T set is the temperature deviation (T set is the safety threshold, T meas (t) is the battery surface temperature measured in real - time by the multi - physical - field sensing layer), K p is the proportional gain, K i is the integral gain, K d .

[0136] In some embodiments, the FPGA chip is built - in with a dual - redundant logic unit. The output results of the main unit and the backup unit are compared every 50 ms. When the difference exceeds 5%, switch to the backup channel and synchronously report the fault code to the cloud platform.

[0137] Please refer to the appendix Figure 4, the active safety execution layer, receives instructions from the edge computing control layer and performs cooling control and circuit protection operations;

[0138] In this embodiment, the phase change cooling unit consists of a microchannel jet array and a phase change material storage tank. The inner diameter of the microchannels is 0.5 - 1.0 mm, and a finger-like layout is adopted to cover the surface of the battery cells. Generally, the working fluid in the storage tank is fluorinated liquid HFE - 7100, with a latent heat of phase change of 160 kJ / kg and a boiling point of 56 °C (matching the safety threshold temperature in the claims).

[0139] Specifically, the cooling power optimization objective function is defined as:

[0140] J = αQ 2 (t) + βΔT 2 (t);

[0141] where Q(t) is the real-time flow rate, with a value range of 0 - 10 L / min, α is the power consumption proportionality coefficient, with a value range of 0.1 - 0.3 (preferred value 0.2), β is the temperature deviation penalty coefficient, with a value range of 5 - 10 (preferred value 8), and e(t) = T meas (t) - T set is the temperature deviation.

[0142] In a possible implementation, the control instruction drives a piezoelectric ceramic pump through a pulse width modulation (PWM) signal, with a duty cycle adjustment accuracy of 1%, meeting the requirement of a flow rate resolution of 0.1 L / min.

[0143] In some embodiments, the phase change material storage tank integrates a thermoelectric power generation module, which uses the heat released by the phase change of the working fluid to generate auxiliary electric energy, with a power generation efficiency of 3%, reducing the net power consumption of the system. In other embodiments, the surface of the microchannels is coated with a nanoporous coating, increasing the boiling heat transfer coefficient to 8000 W / (m 2 ·K), a 40% increase compared to ordinary copper substrates.

[0144] In this embodiment, the magnetohydrodynamic circuit breaker unit adopts a gallium-based liquid metal contact design. Under normal conditions, the liquid metal forms a conductive path under the constraint of a magnetic field. When receiving a secondary alarm signal (Γ(t) ≥ 2500), the electromagnetic coil demagnetizes within 5 ms, and the liquid metal contracts and breaks due to surface tension, achieving physical isolation of the charge and discharge circuit.

[0145] Performance parameters:

[0146] Rated cut-off current: 150 A (matching the maximum output current of the energy storage system);

[0147] Arc suppression ability: The cavity is filled with sulfur hexafluoride, and the breakdown voltage > 10 kV;

[0148] Recovery time: After the fault is removed, it is necessary to manually inject liquid metal and re-excite.

[0149] In some embodiments, the dual-redundancy magnetic field coils are arranged orthogonally, and the standby coil can be activated within 10 ms when the main coil fails. In other embodiments, the contact state monitoring module detects the change of magnetic flux through a Hall sensor and real-time feeds back the signal of the completion of the open-circuit action to the cloud platform.

[0150] In this embodiment, the self-destruction pressure relief unit adopts a Ni-Ti shape memory alloy valve, and its phase transition temperature is set at 85 °C (lower than the critical temperature of 100 °C for the thermal runaway of the battery cell). When the internal pressure of the battery exceeds 2 kPa to 5 kPa, the valve body undergoes an austenite phase transition when heated, and the opening angle reaches 120°, and the effective flow area of the pressure relief channel expands to 150 mm 2 。

[0151] Specifically, the pressure relief channel has a spiral gradually expanding structure, and the gas flow velocity decreases from the initial 30 m / s to 5 m / s to avoid the high-speed gas flow from igniting the escaping combustible gas. The directional pressure relief port faces the preset safe area and forms a 45° isolation angle with the personnel activity area.

[0152] In some embodiments, the surface of the valve is coated with a ceramic heat insulation layer to withstand the impact of instantaneous high-temperature gas (>300 °C). In other embodiments, the pressure sensor array and the temperature data are fused for judgment, and when the pressure rising rate exceeds 500 Pa / s, the pressure relief action is triggered in advance.

[0153] Description of the relevance of the technical solutions:

[0154] Instruction connection mechanism:

[0155] The first-level alarm (Γ(t)≥1500) triggers the PID control of the phase change cooling unit;

[0156] The second-level alarm (Γ(t)≥2500) synchronously activates the magnetohydrodynamic open circuit and the pressure relief unit to form a double protection.

[0157] Global consistency of parameters:

[0158] The safety threshold temperature (45 °C) of the cooling unit corresponds strictly to the set value of the edge computing layer;

[0159] The pressure relief pressure range (2 - 5 kPa) matches the deformation detection sensitivity (30 - 100 μm / m) of the multi-physical field perception layer, satisfying σ max ≥σ crit The associated trigger condition.

[0160] Fail-safe design:

[0161] When the cooling unit loses power, the spring energy storage mechanism pushes the slide valve to the fully open position (10 L / min);

[0162] The magnetohydrodynamic circuit breaker unit is built-in with a fusing fuse, which is forced to blow when the liquid metal fails.

[0163] Please refer to the appendix Figure 5 , the cloud intelligent platform layer, communicates bidirectionally with the edge computing control layer to achieve remote status monitoring and policy optimization.

[0164] In this embodiment, the digital twin unit constructs a three-dimensional transient simulation model through a finite element solver based on the discrete physical field data (temperature, electric potential, stress) uploaded by the edge computing layer. Generally, the finite element mesh is divided into tetrahedral elements, the initial mesh size is 5 mm, and the number of nodes is not less than 10 5 to meet the model accuracy requirements.

[0165] Specifically, the coupling relationship between the transient heat conduction equation and the electric potential field is:

[0166]

[0167] where ρ is the density of the battery cell material (lithium iron phosphate), C p is the specific heat capacity, σ is the electrical conductivity (consistent with the parameters of the electric potential detection unit), is the magnitude of the electric potential gradient (V / m), provided by the reconstructed data of the edge computing layer, Q ext is the external heat source term (such as environmental radiation), and is default set to 0.

[0168] The surface heat transfer boundary condition is expressed as:

[0169]

[0170] where h is the surface convective heat transfer coefficient, T amb is the environmental temperature, obtained in real time by the meteorological data interface, is the spatial gradient of the temperature in the surface normal direction (outer normal direction), with the unit of K / m.

[0171] In some embodiments, a dynamic mesh optimization algorithm is introduced: when the local temperature gradient / cm is detected, the mesh encryption operation is triggered, and the target element size is reduced to 1 mm. The encryption criterion is:

[0172]

[0173] where Δx is the current mesh size, and T avg is the regional average temperature.

[0174] In some other embodiments, an adaptive time step control is adopted, and the CFL condition constraint is:

[0175]

[0176] where α = k / (ρC p ) = 6.25×10 -6 m 2 / s is the thermal diffusivity to ensure numerical stability. In this embodiment, the input layer of the LSTM neural network receives the following time-series features:

[0177] Thermal risk index sequence: Γ(t), sampling interval 10 seconds, time window 24 hours (a total of 8640 data points);

[0178] Capacity attenuation rate: The calculation method is:

[0179]

[0180] where C nom is the nominal capacity, and C meas (t) is the actually measured capacity by the coulomb counting method, with a sampling interval of 1 hour.

[0181] Network structure and hyperparameters:

[0182] Input layer dimension: 8640(Γ(t)) + 24(η(t)) = 8664;

[0183] Hidden layer: 3 layers of LSTM, with 128 neurons in each layer and a dropout rate of 0.2;

[0184] Output layer: Remaining useful life (RUL), in days;

[0185] Loss function: Weighted mean squared error (WMSE):

[0186]

[0187] where the weight enhances the prediction accuracy of long-life batteries, and w i is the weight of the i-th sample, is the predicted remaining useful life of the model for the i-th sample, is the true remaining useful life of the i-th sample.

[0188] In some embodiments, a multi-modal data fusion mechanism is introduced to jointly analyze the voltage relaxation curve and impedance spectrum features. The relaxation time constant is calculated as:

[0189]

[0190] where R ct is the charge transfer resistance, C dl is the double-layer capacitance, and R ohm is the ohmic resistance.

[0191] In some other embodiments, a transfer learning strategy is adopted. The pre-trained model is based on NCA battery data, and the learning rate is set to 1×10 -4 , and the weights of the first two layers of LSTM are frozen.

[0192] In this embodiment, the remote locking instruction trigger requires the following joint authentication:

[0193] Device fingerprint: Generated based on the SHA-256 algorithm, and the input parameters include:

[0194] Battery pack serial number (16 bytes);

[0195] PCB board impedance feature code (8 bytes);

[0196] Production batch hash value (32 bytes).

[0197] Dynamic token: Generated based on the HMAC-TOTP algorithm, with a time window of ±30 seconds and a key update period of 24 hours. The calculation formula is:

[0198] TOTP = Truncate(HMAC-SHA1(K,T)) mod 10 6 ;

[0199] where is the time counter, and t0 is the initial timestamp.

[0200] Geographical location: Integrating GPS (accuracy ±3m) and base station triangulation (accuracy ±50m), the coordinate deviation threshold is set to 50 meters, and the verification formula is:

[0201]

[0202] where x g and y g are the real-time geographical location coordinates of the target device in meters, x b and y b are the preset safety area reference point coordinates in meters, and 50m is the defined allowable position deviation radius. Exceeding this range will trigger a security alarm or locking.

[0203] The locking instruction is encrypted in the SM4-CTR mode, with the initial vector (IV) being a random 16-byte number and the counter step size being 1. The ciphertext encapsulation format includes:

[0204] Version number: 0x01 (national cryptography standard);

[0205] Instruction type: 0x02 (emergency locking);

[0206] Validity period: Unix timestamp (4 bytes);

[0207] Ciphertext payload: AES-GCM encrypted locking parameters.

[0208] In some embodiments, a quantum random number generator (QRNG) is introduced to enhance key security, and the random number entropy value is ≥7.9 bits / byte. In some other embodiments, instruction replay attack protection adopts the combination of an incrementing sequence number and a timestamp hash chain:

[0209] H n = SHA-256(H n-1 ∥t n );

[0210] where H0 is the device initialization seed, H n is the 256-bit hash value generated for the nth operation, with the data type being a 32-byte hexadecimal string, H n-1 is the hash value generated for the previous operation, with the data type being the same as H n , and t n is the timestamp for the nth operation, with the data type being an 8-byte Unix timestamp.

[0211] Dual MCU redundant control module, where the main control MCU executes real-time field reconstruction and risk calculation, and the monitoring MCU independently verifies data consistency. When the output deviation exceeds 3% to 8%, it switches to the safe mode to prevent single-point failure;

[0212] In this embodiment, the dual MCU redundant control module consists of a main control MCU (model STM32H743) and a monitoring MCU (model TMS320F28379D). The main control MCU deploys an improved Kriging field reconstruction algorithm and a risk prediction model, and its execution cycle is strictly synchronized with the edge computing layer (200 ms / time), outputting control instructions to the active safety execution layer. The monitoring MCU independently runs a simplified version of the field reconstruction algorithm (based on linear interpolation) and consistency check logic, and the data sampling rate is increased to 1 kHz for quickly capturing transient anomalies.

[0213] Data consistency verification formula:

[0214]

[0215] where S m is the field reconstruction result output by the main control MCU (such as the temperature field peak value, gradient amplitude value, or risk index Γ(t)), S b is the reference value calculated by the monitoring MCU, Δ is the dual-machine output deviation rate, and the threshold is set to 3% to 8% (dynamically adjusted according to the data type).

[0216] Specifically, when the deviation rate Δ > 8% within three consecutive calculation cycles, the monitoring MCU immediately freezes the instruction output of the main control MCU and activates the following operations:

[0217] Reset the main control MCU through the hardware watchdog circuit;

[0218] Switch to the backup control strategy:

[0219] Lock the flow rate of the phase change cooling unit at 5 L / min;

[0220] The magneto - fluid breaker unit enters the pre - charge state;

[0221] Send a fault code (from 0xE001 to 0xE00F) to the cloud platform, including the abnormal data snapshot and environmental parameters.

[0222] In a possible implementation, the deviation threshold is dynamically adjusted according to the system load:

[0223] When the battery pack temperature T avg > 50 °C, the threshold is tightened to 5%;

[0224] When the risk index Γ(t) > 2000, the threshold is relaxed to 10% (to avoid false switching caused by over - sensitivity).

[0225] In some embodiments, the monitoring MCU integrates an abnormal mode self - learning function to generate a whitelist parameter range through training with historical data:

[0226]

[0227] Among them, μ is the mean of the reference values, σ is the standard deviation, and the statistical period is 30 days, is the effective reference value range.

[0228] In some other embodiments, a hardware redundancy arbiter (implemented by FPGA) is introduced to perform a two - out - of - three voting on the outputs of the dual MCUs:

[0229]

[0230] Among them, V a is the preset safety value of the arbiter, which is linked to the pressure threshold of the pressure relief unit. V final is the final control instruction value output by the arbiter. V m is the control instruction value generated by the main control MCU (STM32H743). V b is the control instruction value independently verified by the monitoring MCU (TMS320F28379D).

[0231] Emergency power supply unit, configured with a supercapacitor energy storage device, which can maintain the operation of sensors, communication modules, and key control circuits for at least 15 minutes after the main power supply is interrupted, and is used to ensure continuous monitoring of the system in an emergency state.

[0232] In this embodiment, the emergency power supply unit is composed of two series-connected carbon-based supercapacitor modules, with a single-group nominal capacity of 500F, a working voltage range of 0 - 16V (extendable to 48V), and an equivalent series resistance (ESR) ≤ 2mΩ. The supercapacitors are coupled to the main power bus through a bidirectional Buck-Boost circuit, and are maintained in a fully charged state by floating charging from the main power supply under normal conditions.

[0233] Capacitance capacity calculation formula:

[0234]

[0235] Where, P load is the total power of emergency loads (including sensors, communication modules, and dual MCU modules), t backup is the minimum power supply duration, V min is the initial voltage of the supercapacitor, V cutoff is the discharge cut-off voltage (to avoid over-discharging of the capacitor).

[0236] If, P load = 50W, t backup = 900s (15 minutes), V min = 48V, V cutoff = 24V.

[0237] Substituting into the calculation gives C req = 1000F, and two 500F capacitors are actually configured in series to meet the capacity requirements.

[0238] Specifically, the power supply management chip (model LTC3350) monitors the change of load current in real time, adjusts the duty cycle of the DC / DC converter through an adaptive PID algorithm, and maintains the bus voltage fluctuation ≤ 5%. The adjustment formula is:

[0239]

[0240] Where:

[0241] V err = V set - V bus : Voltage deviation (V set = 48V);

[0242] K p = 0.5, K i = 0.1, K d= 0.02: PID coefficient (shared with the control parameters of the phase change cooling unit); D: PWM duty cycle (0 - 100%).

[0243] In some embodiments, a photovoltaic charging interface is introduced as an emergency power supply supplement. The peak power of the solar panel is 100W, and the MPPT efficiency is ≥ 97%. The charging control logic is as follows:

[0244]

[0245] Among them:

[0246] η = 0.95: Conversion efficiency;

[0247] V max = 52V: Maximum withstand voltage of the capacitor;

[0248] P pv : Real-time photovoltaic power, provided by the meteorological data interface for irradiance prediction.

[0249] In other embodiments, an integrated lithium battery is used as a secondary backup. The capacity configuration formula is as follows:

[0250]

[0251] Among them, t extend = 30min is the extended power supply duration, DOD = 80% is the depth of discharge, and it is calculated that E bat = 2250Wh.

[0252] Working principle: After the system of this application is started, the flexible sensor array in the multi-physical field perception layer continuously captures the surface temperature distribution of the battery cells. The thin-film voltage sensor synchronously scans the potential gradient, and the micro-strain device detects the mechanical stress change in real time. After the data preprocessing module filters out the environmental noise, the discrete signals are subjected to three-dimensional field reconstruction through the edge computing layer. The Kriging algorithm dynamically assigns weights based on the sensor spacing to generate a continuous temperature field model. The risk prediction unit fuses the time derivative and spatial gradient features to quantify the probability of thermal runaway.

[0253] When the risk index breaks through the preset threshold, the decision control unit sends a hierarchical instruction to the execution layer. The phase change cooling unit adjusts the flow rate according to the quadratic optimization function, and the flow rate of the working fluid in the microchannel changes dynamically with the temperature deviation. The magneto-fluid circuit breaker cuts off the abnormal current circuit within a millisecond window, and the shape memory alloy valve opens the pressure relief path when the pressure exceeds the limit. The physical field snapshots and device state parameters are encrypted and uploaded to the cloud, and the digital twin iteratively updates the thermal-electric coupling field through the finite element solver, and the real-time rendering module maps the pseudo-color map of the temperature distribution.

[0254] The dual-MCU architecture processes control instructions in parallel, and the difference verification mechanism continuously compares the output results. When the main power supply is interrupted, the supercapacitor bank maintains the power supply for key modules to ensure the stability of data acquisition and communication links. The system periodically self-checks the status of each unit, and abnormal events trigger the collaborative response of local execution and cloud strategies. The historical data stream drives the LSTM model to predict the life decay trend, and the adaptive algorithm dynamically corrects the control parameters according to the changes in working conditions.

[0255] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety intelligent management system for a mobile energy storage power source at a construction site, characterized in that, Comprising: A multi-physical field sensing layer for real-time collection of temperature field, electric potential field and mechanical stress field data of the battery pack; An edge computing control layer, communicatively connected to the multi-physical field sensing layer, for reconstructing the physical field distribution and calculating the thermal runaway risk index; An active safety execution layer, receiving instructions from the edge computing control layer and performing cooling control and circuit protection operations; A cloud intelligent platform layer, communicating bidirectionally with the edge computing control layer to achieve remote status monitoring and strategy optimization.

2. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 1, characterized in that, The multi-physical field sensing layer includes: A temperature sensing unit composed of a flexible sensor array, with at least 15 temperature detection points arranged in an icosahedron topology on the surface of each battery cell, the detection accuracy being ±0.5 °C, for generating three-dimensional temperature field distribution data; An electric potential detection unit containing 3 to 8 thin-film voltage sensors, distributed along the diagonal or in a grid pattern of the battery cells, for measuring the electric potential gradient and identifying local overvoltage regions; A deformation sensing unit using 5 to 10 MEMS micro-strain sensors, embedded in the battery housing in a cross layout, for detecting stress anomalies caused by mechanical extrusion or deformation, with a sensitivity of 30 to 100 μm / m.

3. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 2, characterized in that, The temperature sensing unit includes a data preprocessing module, which includes: A 24-bit high-precision analog-to-digital converter for converting analog signals into digital signals, with a sampling frequency not less than 1 kHz; A digital filtering unit configured to suppress noise above 50 Hz through a Butterworth low-pass filter, with a passband attenuation less than 3 dB.

4. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 1, wherein, The edge computing control layer includes: A field reconstruction unit that converts discrete sensor data into a continuous temperature field based on the Kriging spatial interpolation algorithm, the weight coefficient being calculated by the Gaussian function of the sensor spacing, and the correlation length being set to 10 mm to 20 mm, for accurately restoring the thermal field distribution; A risk prediction unit that calculates the thermal runaway risk index through integral operation, the index including an integral term of the square of the temperature change rate and an integral term of the square of the temperature gradient, the two weighting coefficients being 0.2 to 0.5 and 0.8 to 1.5 respectively, for quantifying the thermal runaway risk level; A decision control unit that triggers cooling power adjustment instructions and emergency power-off signals according to the risk index classification.

5. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 4, characterized in that, The field reconstruction unit includes: An FPGA parallel computing architecture for accelerating the update of the three-dimensional temperature field, with a computing cycle not exceeding 200 ms; An adaptive grid division module that automatically increases the local grid density to 1 mm accuracy when the detected temperature gradient exceeds 3 °C / cm to 8 °C / cm, for capturing the details of heat propagation.

6. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 1, characterized in that, The active safety execution layer includes: A phase change cooling unit composed of a microchannel injector and a phase change material storage tank, with a flow control range of 0 to 10 L / min and a response time not exceeding 50 ms, for dynamically adjusting the cooling intensity according to the risk level; A magneto-fluid circuit breaker unit using a liquid metal contact design to cut off the charge and discharge circuit above 150 A within 20 ms after receiving the secondary alarm signal, for preventing the spread of thermal runaway; A self-destructive pressure relief unit configured with a shape memory alloy valve that directionally opens the pressure relief channel when the internal pressure exceeds 2 kPa to 5 kPa, for quickly releasing the internal gas pressure.

7. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 6, characterized in that, The control strategy of the phase change cooling unit satisfies the following: The cooling power consumption has a quadratic function relationship with the flow rate, and the proportionality coefficient is 0.1 to 0.3, which is used to optimize the energy consumption; The temperature deviation penalty term is proportional to the square of the difference between the measured temperature and the safety threshold, and the penalty coefficient is 5 to 10, which is used to balance the cooling efficiency and temperature stability.

8. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 1, wherein, The cloud intelligent platform layer includes: Digital twin unit, used to build a three-dimensional simulation model based on the physical field data uploaded by the edge layer, with a data update cycle of no more than 10 seconds and a surface convective heat transfer coefficient of 8 W / (m 2 ·K) to 15 W / (m 2 ·K); A life prediction unit that analyzes the historical thermal risk index sequence and capacity attenuation rate through an LSTM neural network, and the error in predicting the remaining service life does not exceed 5%; A remote locking unit that adopts a three-factor authentication mechanism of device fingerprint, dynamic token, and geographical location, and sends an encrypted locking instruction after successful verification.

9. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 8, characterized in that, The implementation method of the digital twin unit is as follows: A finite element solver is used to solve the thermoelectric coupling equation, and the boundary conditions include electrode insulation and surface convective heat transfer; A real-time rendering module maps the temperature field data into a pseudo-color three-dimensional model, and the color scale resolution is 0.3°C to 1.0°C, which is used for visual monitoring.

10. The safety intelligent management system of a mobile energy storage power supply at a construction site according to claim 1, characterized in that, It also includes: A dual MCU redundant control module. The main control MCU performs real-time field reconstruction and risk calculation, and the monitoring MCU independently verifies data consistency. When the output deviation exceeds 3% to 8%, it switches to the safe mode to prevent single-point failure; An emergency power supply unit is configured with a supercapacitor energy storage device to maintain the operation of sensors, communication modules, and key control circuits for at least 15 minutes after the main power supply is interrupted, which is used to ensure continuous monitoring in the emergency state of the system.

Images