Method for detecting residual life of battery of solar street lamp

Abstract

The invention provides a solar street lamp battery residual life detection method, and belongs to the technical field of solar street lamps, and the method comprises the steps: carrying out the comprehensive parameter collection and analysis of a solar street lamp battery, firstly obtaining the basic and operation parameters of the battery, carrying out the discharge capacity test and internal resistance measurement, and collecting the operation historical data of the battery; based on the data, a battery feature vector is constructed, a life prediction model is established by using a neural network, and a sample battery is trained. In practical application, real-time parameters, such as voltage, current, surface temperature and accumulative operation duration, of a to-be-detected battery are rapidly detected, and the characteristics are input into the pre-training model, so that the residual life of the battery can be accurately predicted, a scientific and accurate decision basis is provided for maintenance and replacement of the solar street lamp battery, and the service life of the battery is prolonged. And the intelligent level of battery management and the reliability of the system are effectively improved. The technical problem that the reliability is not high when the residual life of the solar street lamp battery is judged according to the use time is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of solar street lamps, and in particular, relates to a method for detecting the remaining life of a solar street lamp battery. Background Art

[0002] As a new type of lighting equipment that uses solar energy to generate electricity, solar street lights are widely popular in urban road lighting and rural area lighting due to their environmental protection and energy-saving characteristics. Compared with traditional street lights, the power supply mode of solar street lights has changed from grid power supply to independent solar cell power supply, which greatly reduces the operation and maintenance costs. However, the service life of the core component of solar street lights - solar cells - is a key factor affecting the performance and working reliability of the entire system.

[0003] At present, the service life of solar street light batteries is generally short, with an average of only 5-8 years, which seriously restricts the large-scale promotion and application of solar street lights. The main reasons for the short battery life include: battery capacity attenuation, increased internal resistance, cycle life exhaustion, etc. If the battery life is too short, it will inevitably increase the replacement frequency of solar street lights, thereby increasing operation and maintenance costs and weakening the economic advantages of solar street lights. Therefore, how to effectively extend the service life of solar street light batteries has become a key technical problem that needs to be solved urgently.

[0004] In response to the above problems, the industry generally simply judges the remaining life of solar street light batteries based on the length of use, which is not very reliable. Summary of the invention

[0005] In view of this, the present invention provides a method for detecting the remaining life of a solar street light battery, which can solve the technical problem that the prior art often simply determines the remaining life of a solar street light battery based on the usage time, and the reliability is not high.

[0006] The present invention is achieved in that: The present invention provides a method for detecting the remaining life of a solar street lamp battery, comprising the following steps: S01. Obtain the basic parameters of the sample solar street light battery, including the battery rated capacity value, battery operating voltage value, battery design life value, and battery internal structure parameter value; S02, collecting the real-time operating parameters of the sample solar street light battery, including the real-time voltage value of the battery, the real-time current value of the battery, the ambient temperature value of the battery, the surface temperature value of the battery, the charging time value of the battery, and the discharging time value of the battery; S03, perform a battery discharge capacity test to measure the actual discharge capacity value of the sample solar street light battery under standard working conditions; S04, continuously acquiring the ratio of the real-time voltage value of the battery to the real-time current value of the battery within a preset sampling period to obtain a battery internal resistance measurement sequence; S05, performing compensation processing on the battery internal resistance measurement sequence based on the temperature compensation equation group to obtain a battery internal resistance compensation value at a standard temperature; S06. Obtain the operation history data of the sample solar street light battery, including the number of battery charge and discharge cycles, the cumulative charge value of the battery, the cumulative discharge value of the battery, the historical maximum temperature value of the battery, and the historical minimum temperature value of the battery; S07, calculating the battery coulombic efficiency value according to the battery cumulative charge value and the battery cumulative discharge value; S08, constructing a battery characteristic vector by using the battery internal resistance compensation value, the battery coulomb efficiency value, the actual discharge capacity value, and the battery charge and discharge cycle number value; S09, establishing a battery life neural network prediction model, inputting the battery feature vector and the corresponding actual remaining life value into the battery life neural network prediction model for training, and obtaining a trained battery life prediction model; S10, collecting rapid detection parameters of the solar street light battery to be tested, including the real-time voltage value of the battery to be tested, the real-time current value of the battery to be tested, and the surface temperature value of the battery to be tested; S11, calculating the ratio of the real-time voltage value of the battery to be tested to the real-time current value of the battery to be tested to obtain an estimated value of the internal resistance of the battery to be tested; S12, obtaining the cumulative running time value of the solar street light battery to be tested; S13, inputting the estimated value of the internal resistance of the battery to be tested, the surface temperature value of the battery to be tested, and the accumulated operating time value into the trained battery life prediction model to obtain a predicted value of the remaining life of the battery to be tested.

[0007] On the basis of the above technical solution, the remaining life detection method of a solar street lamp battery of the present invention can also be improved as follows: The temperature compensation equation group includes a heat conduction equation, a temperature distribution equation, an internal resistance compensation equation, a heat balance equation, and a time response equation.

[0008] Furthermore, the heat conduction equation is used to calculate the temperature conduction relationship between the inside and the surface of the battery, the input includes the battery surface temperature value, the battery ambient temperature value, the battery internal structure parameter value, and the battery real-time current value, and the output is the battery internal temperature gradient value.

[0009] Furthermore, the temperature distribution equation is used to describe the temperature distribution characteristics of each layer structure of the battery. The input includes the internal temperature gradient value of the battery, the internal structure parameter value of the battery, the real-time current value of the battery, the charging time value of the battery, and the discharging time value of the battery. The output is the battery temperature distribution data.

[0010] Furthermore, the internal resistance compensation equation is used to establish a mapping relationship between internal resistance and temperature, and the input includes the battery temperature distribution data, the battery internal resistance measurement sequence, the battery historical maximum temperature value, and the battery historical minimum temperature value, and the output is the battery internal resistance temperature compensation coefficient value.

[0011] Furthermore, the thermal balance equation is used to calculate the steady-state operating temperature of the battery, the input includes the battery ambient temperature value, the battery real-time current value, the battery real-time voltage value, and the battery surface temperature value, and the output is the battery balance temperature value.

[0012] Furthermore, the time response equation is used to describe the dynamic characteristics of temperature changes, and the input includes the battery balance temperature value, the preset sampling period, the battery real-time current value, and the battery real-time voltage value, and the output is the battery dynamic temperature response data.

[0013] Furthermore, the sample solar street lights include at least 100 solar street lights.

[0014] The rapid detection parameters of the solar street light battery to be tested are collected by a signal collection device, and the signal collection device includes: A signal acquisition unit, the signal acquisition unit comprising: A fast voltage acquisition sensor, wherein the fast voltage acquisition sensor is connected in parallel with the solar street light battery to be tested; A fast current acquisition sensor, which is connected in series in the power supply circuit of the solar street light battery to be tested; A rapid temperature acquisition sensor, wherein the rapid temperature acquisition sensor is fixed on the surface of the solar street light battery to be tested; A signal conditioning unit, the signal conditioning unit comprising: An operational amplifier circuit, wherein the operational amplifier circuit is used to amplify the analog signal output by the second signal acquisition unit; A bandpass filter circuit, wherein the bandpass filter circuit is used to filter the amplified signal; A high-speed analog-to-digital conversion circuit, wherein the high-speed analog-to-digital conversion circuit is used to convert the processed analog signal into a digital signal; A control processing unit, the control processing unit comprising: A low-power single-chip microcomputer circuit, wherein the low-power single-chip microcomputer circuit is used to receive the digital signal output by the second signal conditioning unit; A fast storage circuit, the fast storage circuit is connected to the low power consumption single chip microcomputer circuit; A wireless communication circuit is connected to the low-power single-chip microcomputer circuit.

[0015] Furthermore, the battery life neural network prediction model adopts an improved bidirectional long short-term memory network structure, specifically, an attention allocator is added between the memory unit module and the forget gate module in the bidirectional long short-term memory network structure, and the attention allocator is used to assign different weight coefficients to different characteristic parameters in the input battery feature vector, thereby highlighting the key parameters that have a greater impact on the remaining life prediction results; the specific structure of the attention allocator is composed of a feedforward neural network, which includes three hidden layers, the first hidden layer is set with 32 neurons, the second hidden layer is set with 16 neurons, and the third hidden layer is set with 8 neurons. The attention allocator calculates the attention weight value for each characteristic parameter in the input battery feature vector, and multiplies the calculated attention weight value by the original characteristic parameter to generate a weighted characteristic vector.

[0016] Furthermore, the improved bidirectional long short-term memory network structure is described in detail as follows: the input layer receives the battery feature vector, and after weighted processing by the attention allocator, it is input into the bidirectional long short-term memory layer, the bidirectional long short-term memory layer includes a long short-term memory unit for forward propagation and a long short-term memory unit for reverse propagation, each long short-term memory unit is provided with 64 neurons, the outputs of the long short-term memory unit for forward propagation and the long short-term memory unit for reverse propagation are fused through a fully connected layer, the fully connected layer is provided with 128 neurons, batch normalization processing is used to reduce the offset of data distribution, and finally the remaining battery life prediction value is obtained through the output layer, and the output layer adopts a linear activation function.

[0017] Further, the relevant equations or calculations are described in detail as follows: 1. The calculation of the battery internal resistance measurement sequence is specifically expressed as follows: ; In the formula, is the internal resistance measurement value at time t, in ohms; is the real-time voltage value of the battery at time t, in volts; It is the real-time current value of the battery at time t, in amperes.

[0018] 2. The heat conduction equation is specifically expressed as follows: ; In the formula, is the temperature field function, in Kelvin; is the time in seconds; is the thermal diffusion coefficient in square meters per second; is the spatial coordinate; is the heat source term, the pyrolysis caused by the current, in watts per cubic meter; is the specific heat capacity, expressed in joules per kilogram Kelvin; is the density in kilograms per cubic meter.

[0019] 3. The temperature distribution equation is specifically expressed as follows: ; In the formula, is the temperature at time t at a distance r from the center of the battery, in Kelvin; is the battery surface temperature in Kelvin; is the intensity of the heat source, in watts; is the thermal conductivity in Watts per meter Kelvin; is the error function; is the radial distance in meters.

[0020] 4. The internal resistance compensation equation is specifically expressed as follows: ; In the formula, is the internal resistance value after temperature compensation, in ohms; To measure the internal resistance, the unit is ohm; is the temperature coefficient; is the current temperature in Kelvin; is the reference temperature in Kelvin; is the state of charge correction function; Correction function for the number of cycles.

[0021] 5. The heat balance equation is specifically expressed as follows: ; In the formula, is the battery mass in kilograms; is the specific heat capacity, expressed in joules per kilogram Kelvin; is the current in amperes; is the internal resistance in ohms; is the convective heat transfer coefficient in Watts per square meter Kelvin; is the surface area in square meters; is the ambient temperature in Kelvin; is the battery temperature in Kelvin; is the Stefan-Boltzmann constant; is the radiation coefficient.

[0022] 6. The calculation of the battery coulombic efficiency value is specifically expressed as follows: ; In the formula, is the Coulomb efficiency, expressed in percentage; is the discharge current in amperes; is the charging current in amperes; is the discharge start and end time, in seconds; It is the charging start and end time in seconds.

[0023] 7. The time response equation is specifically expressed as follows: ; In the formula, is the battery temperature at time t, in Kelvin; is the system time constant in seconds; is the gain coefficient; is the ambient temperature in Kelvin; is the input power in watts; is the battery mass in kilograms; is the specific heat capacity in joules per kilogram Kelvin.

[0024] 8. The battery feature vector construction is specifically expressed as follows: ; In the formula, is the internal resistance after compensation, in ohms; is the Coulomb efficiency, expressed in percentage; is the actual discharge capacity value in ampere-hours; is the number of charge and discharge cycles; is the health status value, in percentage; is the highest temperature ever recorded, in Kelvin; is the lowest temperature ever recorded, in Kelvin; is the depth of discharge in percentage.

[0025] 9. The attention weight calculation is specifically expressed as follows: ; ; In the formula, is the attention weight of the i-th feature; is the energy score of the i-th feature; is the attention vector; is the weight matrix; is the input feature; is the bias term; is the number of features.

[0026] 10. The forward propagation calculation of the improved bidirectional long short-term memory network is specifically expressed as follows: ; ; ; ; ; ; In the formula, Output of the forget gate; is the input gate output; is a candidate memory unit; is the current memory unit; Output of the output gate; is a hidden state; is the weight matrix; is the bias term; is the current input; is the sigmoid activation function.

[0027] The construction principles and significance of these equations are explained in detail below: 1.Battery internal resistance measurement equation: This equation is constructed based on Ohm's law and uses the voltage-current ratio to calculate the internal resistance. Taking into account the dynamic characteristics of the battery's internal resistance, real-time sampling is used to obtain the instantaneous internal resistance value. This method is simple, effective and easy to implement. The ratio of voltage to current reflects the internal impedance characteristics of the battery. The larger the internal resistance, the worse the battery performance. Continuous sampling can be used to obtain the trend of the internal resistance over time.

[0028] 2. Heat conduction equation: This equation is constructed based on Fourier's law of heat conduction. It uses a three-dimensional heat diffusion model to describe the temperature field distribution inside the battery and introduces a heat source term to consider the effect of focal dissipation caused by current. The partial differential term in the equation describes the propagation characteristics of temperature in space, and the heat source term describes the effect of internal heat on the temperature field. This equation can accurately describe the temperature distribution law inside the battery.

[0029] 3. Temperature distribution equation: The equation is constructed based on the point source instantaneous solution and error function, taking into account the cylindrical structure characteristics of the battery and describing the temperature distribution through radial distance. The introduction of the error function enables the equation to describe the transient conduction process of temperature. The thermal conductivity and thermal diffusion coefficient in the equation reflect the thermal conductivity characteristics of the material. This equation form is easy to solve and has a clear physical meaning.

[0030] 4. Internal resistance compensation equation: This equation takes into account the combined effects of temperature, state of charge and number of cycles on internal resistance, uses a quadratic polynomial to describe the temperature effect, and introduces a correction function to consider other factors. The temperature term uses a quadratic form because the battery internal resistance has a nonlinear relationship with temperature. The introduction of the correction function enhances the adaptability of the equation and makes the internal resistance compensation more accurate.

[0031] 5. Heat balance equation: The equation is constructed based on the principle of conservation of energy and takes into account three heat transfer modes: foci dissipation, convection heat transfer and radiation heat transfer. The left side of the equation describes the temperature change rate of the battery, and the three terms on the right side correspond to three forms of heat transfer. The foci dissipation term is proportional to the square of the current, reflecting the significant influence of large current on temperature rise. The convection term and radiation term describe the heat exchange process with the environment.

[0032] 6. Coulomb efficiency calculation equation: This equation calculates the coulombic efficiency through the time-integrated ratio of the charge and discharge currents. The integral form can accurately reflect the energy conversion efficiency of the entire charge and discharge process. The efficiency value directly reflects the energy conversion performance of the battery and is an important indicator for evaluating battery performance.

[0033] 7. Time response equation: The equation uses a first-order differential equation to describe the dynamic response characteristics of the battery temperature, taking into account the influence of system inertia and external excitation; the time constant reflects the response speed of the system, the gain coefficient reflects the amplification effect of the system, and the power term describes the influence of internal heating.

[0034] 8. Battery feature vector construction: The vector design uses multi-dimensional features to characterize the battery status, including electrochemical properties (internal resistance, coulombic efficiency), capacity characteristics (actual discharge capacity), usage history (number of cycles), health status (SOH) and temperature characteristics (maximum and minimum temperatures) and usage intensity (depth of discharge); the selection of each feature is based on its impact on the battery life, forming a comprehensive feature space, which is convenient for subsequent neural network training and prediction.

[0035] 9. Attention weight calculation equation: This equation is constructed based on the theory of attention mechanism. The feature weights are calculated by energy scores to achieve adaptive adjustment of the importance of different features. The exponential softmax ensures that the sum of the weights is 1 and they are all positive values. The tanh activation function introduces nonlinear characteristics to enhance the expressiveness of the model. The entire mechanism can highlight the role of key features.

[0036] 10. Improved bidirectional long short-term memory network calculation equations: This set of equations describes the core computational process of LSTM, including the calculation of forget gate, input gate, memory cell update, and output gate; The forget gate equation determines the degree of retention of historical information, and the sigmoid function limits the output between 0 and 1; The input gate equation controls the acceptance level of the current input information and jointly determines the new information with the candidate memory unit; The memory unit update equation performs a weighted combination of historical information and new information to achieve modeling of long-term dependencies. The output gate equation controls the output degree of information and passes the processed information to the next moment.

[0037] Compared with the prior art, the method for detecting the remaining life of a solar street lamp battery provided by the present invention has the following beneficial effects: 1. Rapidly and non-destructively acquire key battery parameters. The present invention uses fast voltage, current, and temperature acquisition sensors to acquire key state parameters such as internal resistance and temperature of the battery to be tested in real time and with high precision without disassembling the battery, greatly improving data acquisition efficiency. This provides a reliable input data basis for subsequent remaining life prediction.

[0038] 2. Establish an accurate battery life prediction model. This invention integrates multiple core technologies such as battery internal resistance measurement, temperature compensation, and machine learning to construct an improved bidirectional long short-term memory neural network model that can more accurately predict the remaining service life of the battery. The model automatically learns the importance of various battery characteristic parameters to life prediction through the attention mechanism, greatly improving the accuracy of the prediction results.

[0039] 3. Extend battery life and reduce operation and maintenance costs. Based on the remaining battery life prediction results obtained by the method of the present invention, operation and maintenance personnel can promptly discover signs of battery aging and reasonably arrange battery replacement plans in advance, thereby avoiding premature battery failure, giving full play to the battery life, and reducing the overall maintenance cost of the solar street light system.

[0040] In summary, the present invention solves the technical problem that the prior art often simply determines the remaining life of the solar street light battery based on the usage time, which is not very reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of the method provided by the present invention; Figure 2 This is a schematic diagram of a solar street light battery signal collection device in an embodiment of the present invention; Figure 3 The curve diagram of the change of the internal resistance of the battery with the number of charge and discharge cycles in the embodiment; Figure 4 Schematic diagram of the performance evaluation results of the battery remaining life prediction model. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0043] like Figure 1 As shown, it is a flow chart of a method for detecting the remaining life of a solar street lamp battery provided by the present invention. The specific implementation method of each step of the present invention is described in detail below: The specific implementation method of step S01 is to first obtain the battery rated capacity value through the product specification provided by the battery manufacturer, which is usually 20 ampere hours to 100 ampere hours, the battery operating voltage value is generally 12 volts to 48 volts, and the battery design life value is usually 3 to 5 years. The battery internal structure parameter values ​​include the positive electrode material thickness of 0.1 mm to 0.3 mm, the negative electrode material thickness of 0.08 mm to 0.25 mm, the diaphragm thickness of 0.02 mm to 0.05 mm, and the electrolyte concentration of 1 mol per liter to 2 mol per liter. These basic parameters are verified and confirmed by a dedicated battery parameter tester. The measurement accuracy of the tester should not be less than 0.1%, the test environment temperature should be maintained at around 25 degrees Celsius, and the relative humidity should be controlled between 45% and 65%.

[0044] The specific implementation method of step S02 is to monitor the battery operating status in real time through a high-precision data acquisition system, the sampling frequency is set to 1 Hz, the real-time voltage value of the battery is measured by a Hall voltage sensor, the measuring range is 0 volts to 60 volts, and the measuring accuracy is 0.1%. The real-time current value of the battery is measured by a Hall current sensor, the measuring range is 0 amperes to 50 amperes, and the measuring accuracy is 0.1%. The battery ambient temperature value is measured by a PT100 temperature sensor, the measuring range is minus 40 degrees Celsius to plus 85 degrees Celsius, and the measuring accuracy is 0.1 degrees Celsius. The battery surface temperature value is measured by an infrared temperature sensor, the measuring range is minus 40 degrees Celsius to plus 85 degrees Celsius, and the measuring accuracy is 0.1 degrees Celsius. The battery charging time value and the discharge time value are obtained through system clock timing, and the timing accuracy is 1 second. All collected data are converted into digital signals through an analog-to-digital converter after being processed by a signal conditioning circuit, using 16-bit analog-to-digital conversion accuracy, and the conversion rate is not less than 100 kHz.

[0045] The specific implementation method of step S03 is to use a professional battery testing system to perform a discharge capacity test under standard working conditions of an ambient temperature of 25 degrees Celsius and a relative humidity of 60%. First, the battery is charged to a full charge state, the charging current is 0.2 times the rated capacity value, and the charging cut-off voltage is 2.4 volts for a single cell. Then, the battery is left to stand for 1 hour to allow the battery temperature to stabilize. Then, the battery is discharged at a constant current of 0.2 times the rated capacity value, and the discharge cut-off voltage is 1.8 volts for a single cell. The current integral value recorded during the discharge process is the actual discharge capacity value. During the test, the battery voltage, current and temperature data are recorded every 10 seconds. The entire test process is repeated 3 times and the average value is taken as the final result.

[0046] The specific implementation method of step S04 is to collect the real-time voltage value and the real-time current value of the battery every 1 second within a preset sampling period, and the sampling time is not less than 1 hour. The battery internal resistance measurement value is obtained by dividing the real-time voltage value by the real-time current value to form a battery internal resistance measurement sequence containing 3600 data points. The measurement sequence is smoothed by a sliding average filtering algorithm, and the sliding window length is set to 10 data points to eliminate the influence of measurement noise. At the same time, a median filtering algorithm is used to remove outliers, and the filter window length is set to 5 data points. Finally, the processed data is fitted by the least squares method to obtain the battery internal resistance change trend curve.

[0047] The specific implementation method of step S05 is to construct a temperature compensation model based on the heat conduction equation, the temperature distribution equation, the internal resistance compensation equation, the heat balance equation and the time response equation. First, the temperature conduction relationship between the internal layers of the battery is calculated by the heat conduction equation, and the thermal diffusion coefficient ranges from 1.0×10-6 square meters per second to 5.0×10-6 square meters per second. Then, the radial temperature distribution characteristics of the battery are analyzed by the temperature distribution equation, and the thermal conductivity ranges from 0.5 watts per meter Kelvin to 2.0 watts per meter Kelvin. Then, the internal resistance compensation equation is used to establish a quantitative relationship between the internal resistance and temperature. The temperature coefficient k1 ranges from 0.001 to 0.005, and the k2 ranges from 0.0001 to 0.0005. Then, the steady-state operating temperature of the battery is calculated by the heat balance equation, and the convection heat transfer coefficient ranges from 5 watts per square meter Kelvin to 20 watts per square meter Kelvin. Finally, the time response equation is used to describe the dynamic temperature change characteristics, and the system time constant ranges from 100 seconds to 300 seconds.

[0048] The specific implementation method of step S06 is to extract the battery operation history data from the history database of the battery management system, the battery charge and discharge cycle number value is obtained by detecting the voltage inflection point of the charge and discharge process, and a discharge depth greater than 20% and charging recovery to more than 90% is recorded as one effective cycle, the battery cumulative charge value is obtained by time integration of the charging current, and the integration time interval is 1 second, the battery cumulative discharge value is obtained by time integration of the discharge current, and the integration time interval is 1 second, the battery historical maximum temperature value and minimum temperature value are obtained by screening the temperature monitoring data, and the data query time span is the entire operation cycle of the battery since it was put into use.

[0049] The specific implementation method of step S07 is to calculate the coulombic efficiency value of the battery according to the cumulative charging value and the cumulative discharging value of the battery. First, the charging current and the discharging current are integrated over time to obtain the total charging and discharging capacity. The integration adopts the trapezoidal integration method to improve the calculation accuracy. Then, the percentage of the total discharge capacity to the total charging capacity is calculated to obtain the coulombic efficiency value. The coulombic efficiency value is usually between 85% and 95%. If the coulombic efficiency value is lower than 80%, it indicates that the battery performance has obviously declined and needs to be paid special attention. The influence of the current measurement error needs to be considered in the calculation process. Generally, the relative error of the current measurement is required not to exceed 0.5%.

[0050] The specific implementation method of step S08 is to construct a battery feature vector after standardizing characteristic parameters such as battery internal resistance compensation value, battery coulomb efficiency value, actual discharge capacity value, and battery charge and discharge cycle number value. The standardization adopts the minimum and maximum value normalization method to map each characteristic parameter to the range of 0 to 1, wherein the standardized range of internal resistance compensation value is 0.5 milliohms to 5 milliohms, the standardized range of coulomb efficiency value is 80% to 100%, the standardized range of actual discharge capacity value is 60% to 100% of the rated capacity, and the standardized range of charge and discharge cycle number value is 0 to 2000 times. The dimension of the constructed feature vector is 8 dimensions, which contains the main characteristic information reflecting the battery performance degradation.

[0051] The specific implementation method of step S09 is to use an improved bidirectional long short-term memory network to construct a battery life prediction model. First, a network structure with an attention mechanism is designed. The input layer receives an 8-dimensional battery feature vector. The attention allocator adopts a three-layer feedforward neural network structure with 32 neurons in the first layer, 16 neurons in the second layer, and 8 neurons in the third layer. The activation function uses a hyperbolic tangent function. Then, a bidirectional long short-term memory layer is constructed, which includes 64 memory units in the forward and reverse directions. The network is trained using a stochastic gradient descent algorithm. The learning rate is set to 0.001, the batch size is set to 32, and the number of training rounds is set to 500. Finally, the bidirectional outputs are fused through a fully connected layer. 128 neurons are set, and a batch normalization method is used to reduce data distribution offset.

[0052] The specific implementation method of step S10 is to use a signal acquisition device to perform rapid parameter detection on the solar street light battery to be tested. The rapid voltage acquisition sensor is implemented by a high-precision voltage divider circuit, with a range of 0 volts to 60 volts and a sampling rate of 1 kHz. The rapid current acquisition sensor is implemented by a Hall sensor, with a range of 0 amperes to 50 amperes and a sampling rate of 1 kHz. The rapid temperature acquisition sensor is implemented by a thermocouple, with a range of minus 40 degrees Celsius to plus 85 degrees Celsius and a sampling rate of 10 Hz. The output signals of all sensors are processed by a signal conditioning circuit, including operational amplification, bandpass filtering, and analog-to-digital conversion.

[0053] The specific implementation method of step S11 is to obtain the estimated internal resistance of the battery to be tested by calculating the ratio of the real-time voltage value of the battery to be tested to the real-time current value of the battery to be tested. The acquisition time is not less than 10 minutes, the sampling interval is 1 second, and no less than 600 internal resistance estimation data points are obtained. The estimated data is processed by the Kalman filter algorithm to remove the influence of measurement noise and outliers. The process noise covariance matrix and the measurement noise covariance matrix of the Kalman filter are obtained through experimental calibration. Finally, the mean of the processed data is calculated as the final internal resistance estimation value.

[0054] The specific implementation method of step S12 is to obtain the cumulative operating time value of the solar street light battery to be tested by querying the operation record of the battery management system. The cumulative operating time is calculated from the time the battery is first installed and put into use, accurate to hours. The operation record includes the battery's working status information, such as charging status, discharging status, idle status, etc. By analyzing these status information, the time when the battery is not working can be eliminated to obtain the true cumulative operating time value.

[0055] The specific implementation method of step S13 is to input the estimated value of the internal resistance of the battery to be tested, the surface temperature value of the battery to be tested, and the accumulated operating time value into the trained battery life prediction model, first standardize the input parameters, and use the same normalization method as the training data, and then calculate the weight coefficient of each input feature through the attention allocator, and then input the weighted feature vector into the bidirectional long short-term memory network for forward and reverse propagation calculations, and finally obtain the remaining life prediction value of the battery to be tested through the fully connected layer and the output layer, and the confidence interval of the prediction result is plus or minus 10%.

[0056] Specifically, the principle of the present invention is: first, for the important state parameter of battery internal resistance, the present invention adopts a fast voltage and current acquisition sensor, which can obtain the internal resistance data of the battery in real time and with high precision. Internal resistance is a key indicator reflecting the electrochemical performance of the battery. It will continue to increase as the battery ages, so the internal resistance data is a key input for predicting the remaining life of the battery. In order to eliminate the interference of temperature on the internal resistance measurement, the present invention also introduces a temperature compensation model, establishes the mapping relationship between internal resistance and temperature through the heat conduction equation, temperature distribution equation, etc., and finally obtains the internal resistance value at the standard temperature.

[0057] Secondly, the present invention also obtains other important state parameters of the battery, such as actual discharge capacity, coulomb efficiency, number of charge and discharge cycles, maximum and minimum temperatures, etc., to construct a complete battery feature vector. These parameters fully reflect the battery's usage status and historical operation trajectory, providing the necessary input basis for subsequent remaining life prediction.

[0058] Finally, the present invention adopts an improved bidirectional long short-term memory (Bi-LSTM) neural network as a battery life prediction model. The network adds an attention allocator module based on the standard Bi-LSTM structure, which can automatically identify the importance of various battery characteristic parameters to the remaining life prediction results and assign different weights according to the importance. This weighting mechanism can highlight key parameters and improve the accuracy and generalization ability of the prediction model. Through training with historical data of a large number of sample batteries, the prediction model can eventually accurately predict the remaining service life of the battery to be tested.

[0059] The following is an example of a specific application scenario of the present invention: 200 solar street lights are installed on a certain road section, and each street light is equipped with a solar panel and a lead-acid battery. The road section is located in the suburbs and has good lighting conditions, but due to environmental factors, the battery life is generally short, with an average of only 6-7 years. The operation and maintenance department hopes to accurately predict the remaining service life of each battery, so as to formulate a reasonable replacement plan and reduce maintenance costs. The operation and maintenance personnel decided to use the solar street light battery remaining life detection method proposed in the present invention for a pilot application. First, the operation and maintenance personnel selected 100 of the 200 street lights as samples and conducted detailed parameter collection and performance testing on these 100 batteries. The specific steps are as follows: 1. Obtain the basic parameters of the sample battery For 100 sample batteries, the operation and maintenance personnel recorded basic information such as rated capacity, operating voltage, design life, and internal structural parameters. Taking a certain battery as an example, its basic parameters are shown in Table 1: Table 1 Basic parameters of a sample battery

[0060] 2. Collect real-time operating parameters of sample batteries In the next three months, the operation and maintenance personnel monitored 100 sample batteries every day through the signal acquisition device, and recorded their real-time voltage, real-time current, ambient temperature, battery surface temperature, charging time, and discharge time. Take the monitoring data of one day as an example, as shown in Table 2: Table 2 Real-time operating parameters of 100 sample batteries on a certain day

[0061] 3. Perform battery discharge capacity test After completing 3 months of real-time parameter collection, the operation and maintenance personnel conducted standard discharge capacity tests on 100 sample batteries one by one. Through this test, the actual discharge capacity value of each battery under standard working conditions can be obtained. Taking a certain battery as an example, its actual discharge capacity is 75Ah.

[0062] 4. Measure battery internal resistance and perform temperature compensation Based on the real-time parameter collection, the operation and maintenance personnel also measured the internal resistance of 100 sample batteries. The specific method is to continuously record the real-time voltage and current values ​​of each battery within a preset sampling period (for example, 1 hour), and then calculate its internal resistance value. : ; Since the internal resistance value will fluctuate with temperature, temperature compensation is also required. Based on the temperature compensation equations mentioned above, the operation and maintenance personnel calculated the internal resistance compensation value at standard temperature. : ; in, is the measured internal resistance value, and is the temperature coefficient, is the current temperature, is the reference temperature, and They are the correction functions of the state of charge and the number of cycles, respectively. Through this step of temperature compensation, the interference of ambient temperature on the internal resistance measurement can be eliminated, and more accurate internal resistance data can be obtained. Figure 3 The curves of battery internal resistance changing with the number of charge and discharge cycles at different operating temperatures (10°C, 25°C and 40°C) are shown. As can be seen from the figure, with the increase of the number of charge and discharge cycles, the battery internal resistance shows a nonlinear upward trend, and the higher the temperature, the faster the internal resistance rises. This figure reflects the influence of temperature on the change of battery internal resistance and provides a basis for temperature compensation.

[0063] 5. Obtain the operating history data of the sample battery In addition to real-time parameters, the operation and maintenance personnel also collected historical operation data of 100 sample batteries, including the number of charge and discharge cycles, cumulative charge amount, cumulative discharge amount, historical highest temperature and historical lowest temperature, etc. These data reflect the battery usage status and environmental load, which will provide important basis for subsequent life prediction.

[0064] 6. Calculate the battery coulombic efficiency Based on the battery charge and discharge history data obtained, the operation and maintenance personnel calculated the coulomb efficiency of each sample battery. : ; in, is the discharge current, is the charging current, arrive is the discharge time period, arrive The coulombic efficiency reflects the energy loss of the battery during the charging and discharging process and is an important indicator for judging battery performance.

[0065] 7. Construct battery feature vector The above-obtained battery parameters, including the internal resistance compensation value , Coulomb efficiency , Actual discharge capacity , Charge and discharge cycle times , health status , the highest temperature in history , the lowest temperature in history and depth of discharge , combined into a feature vector : ; This feature vector comprehensively describes the internal state and usage environment of the battery, providing the necessary input for subsequent life prediction.

[0066] 8. Train the battery life prediction model Based on the feature vectors of the above 100 sample batteries Based on the actual remaining life value, the operation and maintenance personnel established an improved bidirectional long short-term memory (Bi-LSTM) neural network model to predict the remaining service life of the battery. The structure of the model is as follows: The input layer receives the feature vector , after weighted processing by the attention distributor, it enters the Bi-LSTM layer. The Bi-LSTM layer includes LSTM units for forward propagation and LSTM units for backward propagation, and each LSTM unit is set with 64 neurons. The output of the Bi-LSTM layer is fused through the fully connected layer, which is set with 128 neurons and uses batch normalization technology to reduce data distribution deviation. Finally, the predicted value of the remaining battery life is obtained through the linear output layer.

[0067] The attention allocator calculates the attention weight of each feature parameter through a 3-layer feedforward neural network. : ; ; in, is the attention vector, is the weight matrix, is the i-th input feature, is the bias term, is the number of features. This weighting mechanism can highlight the key parameters that have a greater impact on the remaining life prediction results.

[0068] After training with a large amount of sample data, the Bi-LSTM prediction model finally achieved satisfactory prediction accuracy.

[0069] 9. Rapidly test the battery to be tested With the above trained battery life prediction model, the operation and maintenance personnel can quickly test the batteries on the remaining 100 street lights. The specific steps are as follows: First, through the signal acquisition device (such as Figure 2 The signal acquisition device (as shown in Figure 1) collects the key parameters of the battery under test in real time, including real-time voltage, real-time current and surface temperature. The signal acquisition device consists of three sensor units, one signal conditioning unit and one control processing unit, which can collect the required data quickly and accurately.

[0070] Then, the operator calculates the estimated internal resistance of the battery to be tested. : ; And get its cumulative running time value.

[0071] Finally, the estimated internal resistance , surface temperature By inputting the accumulated running time into the trained Bi-LSTM prediction model, the remaining life prediction value of the battery to be tested can be obtained.

[0072] Figure 4 The performance evaluation results of the battery remaining life prediction model are shown. The horizontal axis in the figure is the actual remaining life, the vertical axis is the predicted remaining life, the scattered points represent the corresponding relationship between the predicted value and the actual value, and the red dotted line is the ideal prediction line (the predicted value is equal to the actual value). From the distribution of the scattered points, it can be seen that the correlation between the predicted value and the actual value is high, and the prediction error is within an acceptable range.

[0073] 10. Create a battery replacement plan Based on the remaining life prediction results of all 100 sample batteries and 100 test batteries obtained in the above steps, the operation and maintenance personnel formulated an overall plan for battery replacement: First, for those batteries with a predicted remaining life of less than 2 years, replace them immediately. These batteries have entered the final stage of aging and will have a greater risk of failure if they continue to be used.

[0074] Secondly, for batteries with a predicted remaining life of 2-5 years, the replacement time should be appropriately postponed, but inspection and monitoring efforts should be increased to keep an eye on changes in their usage status.

[0075] Finally, for batteries with a predicted remaining life of more than 5 years, they can be used as planned without immediate replacement. However, their performance indicators should be tested regularly to detect signs of aging in time and adjust the replacement plan appropriately.

[0076] This differentiated replacement strategy can not only maximize the battery life, but also effectively control maintenance costs. It is expected that after one year of pilot application, the average service life of solar street light batteries in this section will be increased from 6-7 years to 8-9 years, significantly reducing the frequency of battery replacement and maintenance workload.

[0077] It should be noted that the explanations of the variables involved in the present invention are shown in Table 3 below.

[0078] Table 3 Variable explanation table

[0079] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for detecting the remaining life of a solar street light battery, characterized in that: The following steps are involved: S01. Obtain the basic parameters of the sample solar street light battery, including the battery rated capacity value, battery operating voltage value, battery design life value, and battery internal structure parameter value; S02, collecting the real-time operating parameters of the sample solar street light battery, including the real-time voltage value of the battery, the real-time current value of the battery, the ambient temperature value of the battery, the surface temperature value of the battery, the charging time value of the battery, and the discharging time value of the battery; S03, perform a battery discharge capacity test to measure the actual discharge capacity value of the sample solar street light battery under standard working conditions; S04, continuously acquiring the ratio of the real-time voltage value of the battery to the real-time current value of the battery within a preset sampling period to obtain a battery internal resistance measurement sequence; S05, performing compensation processing on the battery internal resistance measurement sequence based on the temperature compensation equation group to obtain a battery internal resistance compensation value at a standard temperature; S06. Obtain the operation history data of the sample solar street light battery, including the number of battery charge and discharge cycles, the cumulative charge value of the battery, the cumulative discharge value of the battery, the historical maximum temperature value of the battery, and the historical minimum temperature value of the battery; S07, calculating the battery coulombic efficiency value according to the battery cumulative charge value and the battery cumulative discharge value; S08, constructing a battery characteristic vector by using the battery internal resistance compensation value, the battery coulomb efficiency value, the actual discharge capacity value, and the battery charge and discharge cycle number value; S09, establishing a battery life neural network prediction model, inputting the battery feature vector and the corresponding actual remaining life value into the battery life neural network prediction model for training, and obtaining a trained battery life prediction model; S10, collecting rapid detection parameters of the solar street light battery to be tested, including the real-time voltage value of the battery to be tested, the real-time current value of the battery to be tested, and the surface temperature value of the battery to be tested; S11, calculating the ratio of the real-time voltage value of the battery to be tested to the real-time current value of the battery to be tested to obtain an estimated value of the internal resistance of the battery to be tested; S12, obtaining the cumulative running time value of the solar street light battery to be tested; S13, inputting the estimated value of the internal resistance of the battery to be tested, the surface temperature value of the battery to be tested, and the accumulated operating time value into the trained battery life prediction model to obtain a predicted value of the remaining life of the battery to be tested.

2. A method for detecting the remaining life of a solar street light battery according to claim 1, characterized in that: The temperature compensation equation group includes a heat conduction equation, a temperature distribution equation, an internal resistance compensation equation, a heat balance equation, and a time response equation.

3. A method for detecting the remaining life of a solar street light battery according to claim 2, characterized in that: The heat conduction equation is used to calculate the temperature conduction relationship between the inside and the surface of the battery. The input includes the battery surface temperature value, the battery ambient temperature value, the battery internal structure parameter value, and the battery real-time current value. The output is the battery internal temperature gradient value.

4. A method for detecting the remaining life of a solar street light battery according to claim 3, characterized in that: The temperature distribution equation is used to describe the temperature distribution characteristics of each layer structure of the battery. The input includes the internal temperature gradient value of the battery, the internal structure parameter value of the battery, the real-time current value of the battery, the battery charging time value, and the battery discharging time value. The output is the battery temperature distribution data.

5. A method for detecting the remaining life of a solar street light battery according to claim 4, characterized in that: The internal resistance compensation equation is used to establish a mapping relationship between internal resistance and temperature. The input includes the battery temperature distribution data, the battery internal resistance measurement sequence, the battery historical maximum temperature value, and the battery historical minimum temperature value. The output is the battery internal resistance temperature compensation coefficient value.

6. A method for detecting the remaining life of a solar street light battery according to claim 5, characterized in that: The thermal balance equation is used to calculate the steady-state operating temperature of the battery. The input includes the battery ambient temperature value, the battery real-time current value, the battery real-time voltage value, and the battery surface temperature value. The output is the battery balance temperature value.

7. A method for detecting the remaining life of a solar street light battery according to claim 6, characterized in that: The time response equation is used to describe the dynamic characteristics of temperature changes. The input includes the battery balance temperature value, the preset sampling period, the battery real-time current value, and the battery real-time voltage value. The output is the battery dynamic temperature response data.

8. A method for detecting the remaining life of a solar street light battery according to claim 7, characterized in that: The sample solar street lights include at least 100 solar street lights.

9. A method for detecting the remaining life of a solar street light battery according to claim 8, characterized in that: The battery life neural network prediction model adopts an improved bidirectional long short-term memory network structure, specifically, an attention allocator is added between the memory unit module and the forget gate module in the bidirectional long short-term memory network structure, and the attention allocator is used to assign different weight coefficients to different characteristic parameters in the input battery feature vector, so as to highlight the key parameters that have a greater impact on the remaining life prediction results; the specific structure of the attention allocator is composed of a feedforward neural network, which includes three hidden layers, the first hidden layer is set with 32 neurons, the second hidden layer is set with 16 neurons, and the third hidden layer is set with 8 neurons. The attention allocator calculates the attention weight value for each characteristic parameter in the input battery feature vector, and multiplies the calculated attention weight value by the original characteristic parameter to generate a weighted characteristic vector.

10. A method for detecting the remaining life of a solar street light battery according to claim 9, characterized in that: The improved bidirectional long short-term memory network structure is described in detail as follows: the input layer receives the battery feature vector, which is input into the bidirectional long short-term memory layer after weighted processing by the attention allocator. The bidirectional long short-term memory layer includes a long short-term memory unit for forward propagation and a long short-term memory unit for reverse propagation. Each long short-term memory unit is provided with 64 neurons. The outputs of the long short-term memory unit for forward propagation and the long short-term memory unit for reverse propagation are fused through a fully connected layer. The fully connected layer is provided with 128 neurons. Batch normalization processing is used to reduce the offset of data distribution. Finally, the remaining battery life prediction value is obtained through the output layer. The output layer adopts a linear activation function.

Images