Anti-aging temperature control system for LED transparent display screen

Abstract

The application relates to the technical field of display screen regulation, and discloses an anti-aging temperature control system of an LED transparent display screen, which comprises the following: a collection unit that obtains the measured temperature, a control period and heat injection event information of each subarea in real time to form a pulse sequence for calculation; a thermal age centroid calculation unit that obtains a transient thermal impedance weighted thermal age centroid based on the pulse sequence and a transient thermal impedance function; a temperature prediction unit that inputs historical temperature and a thermal age centroid sequence into a network by using a long short-term memory neural network to output a subarea temperature prediction value; a mapping unit that establishes a quantitative relationship between a heat injection timing displacement and amplitude scaling and a thermal age centroid increment; a sensitivity calculation unit that outputs local temperature sensitivity in combination with a neural network after a minimum disturbance is applied; and an adjustment unit that obtains a minimum change timing displacement and amplitude scaling under the condition that the temperature does not exceed a set value constraint according to the temperature prediction value, the sensitivity and the thermal age centroid increment, generates and issues a control schedule.

Description

Technical Field

[0001] This invention relates to the field of display screen control technology, and more specifically, to an anti-aging temperature control system for LED transparent displays. Background Technology

[0002] As a key application of silicon-based display technology, LED transparent displays are typically installed on glass curtain walls, shop windows, or building facades. To maintain transparency, their structure consists of strip-shaped light-emitting modules and silicon-based driving units. The silicon-based driving units, with their high integration and miniaturization, further optimize the high-aperture grid form of the transparent screen, meeting both the precision integration requirements of silicon-based displays and the visual transparency needs of building facades. During operation, this structure simultaneously withstands heat generated by electrical power and solar radiation. Internal heat is dissipated through conduction, convection, and radiation. While silicon-based materials possess certain thermal conductivity advantages, the short heat dissipation path, thinness, and low heat capacity of the transparent screen mean its temperature rise rate is still much higher than that of traditional cabinet screens. Furthermore, the limited air gaps between some modules and the glass easily lead to localized temperature rises and heat accumulation. When the ambient temperature changes abruptly or the brightness switches frequently, the response time of the temperature in each zone differs significantly from the steady-state temperature. The fast response characteristics of silicon-based drives further amplify the instantaneous nature of this temperature fluctuation, requiring the system to have higher timing resolution to predict and control temperature changes.

[0003] Transparent displays typically employ multi-channel scanning driven methods supported by silicon-based driver chips, such as 1 / 8 or 1 / 16 scanning. The high-precision timing control capability of silicon-based chips allows each zone to be precisely and sequentially illuminated within a refresh cycle. Simultaneously, to maintain stable brightness, pulse-width modulation (PWM) dimming is superimposed. While the high-speed switching characteristics of silicon-based drivers improve dimming accuracy, they also cause more drastic changes in the duty cycle per unit time. Although the average power is the same, the more precise control of the conduction phase of each zone under the silicon-based driver architecture results in a systematic shift in the timing of instantaneous current and heat peaks. This difference in heat input caused by silicon-based driver timing creates an alternating distribution of heat generation followed by cooling, making the local temperature change trend more noticeably delayed and non-linear. When external wind speed, ambient temperature, or content brightness changes, the dynamic response characteristics of silicon-based displays cause this delay to become unpredictable, becoming a major reason why traditional linear control methods struggle to accurately model this delay.

[0004] Existing temperature control typically relies on real-time temperature feedback, using simple proportional or PID adjustments to limit power or stop fans. However, in multi-zone scenarios of transparent displays powered by silicon-based displays, the high response speed and high-density zone design of silicon-based drives make temperature changes more significantly affected by the coupling of time-series driving and environmental disturbances. Single feedback control not only cannot predict the upcoming temperature rise in advance but also struggles to cope with the rapid dynamic temperature fluctuations inherent in silicon-based displays. Because the heat release time of different zones is dispersed, traditional methods can only passively respond after the temperature rises, often resulting in localized overheating or excessive overall brightness reduction. To proactively predict these rapid, non-linear dynamic temperature changes unique to silicon-based displays, a model with time memory capabilities needs to be introduced to establish a predictive relationship between temperature and the time-series changes of silicon-based drives. This allows for the pre-calculation and adjustment of the driving parameters for each zone, stabilizing the temperature within the target range. Summary of the Invention

[0005] This invention provides an anti-aging temperature control system for LED transparent displays, solving the technical problem that in LED transparent displays, due to the use of scanning drive and pulse width modulation light emission control, the heat intensity and duration of different zones of the screen are not consistent at the same time. Light emission units with different scanning phases are turned on sequentially within a frame cycle, resulting in deviations in heat release time. Furthermore, the combined effects of ambient temperature, ventilation conditions, and changes in content brightness cause frequent temperature fluctuations and significant dynamic delays in each zone. Traditional temperature control methods rely solely on current temperature or average power adjustment, failing to accurately predict this temperature lag caused by changes in drive timing, thus leading to overshoot, undershoot, and uneven temperature distribution in the control system.

[0006] This invention provides an anti-aging temperature control system for LED transparent displays, comprising:

[0007] The acquisition unit is used to acquire the measured temperature, control cycle, window duration, occurrence time of heat injection event and equivalent energy of each zone, and to form a pulse sequence based on the occurrence time of heat injection event and equivalent energy.

[0008] The thermal age centroid calculation unit is used to calculate the transient thermal impedance weighted thermal age centroid based on the pulse sequence and the transient thermal impedance function.

[0009] The temperature prediction unit is used to output zoned temperature prediction values ​​through a long short-term memory neural network based on the historical sequence of measured temperatures and the historical sequence of transient thermal impedance weighted thermal age centroids.

[0010] The mapping unit is used to establish the mapping relationship between the heat injection time-series displacement and the heat injection amplitude scaling to the transient thermal impedance weighted thermal age centroid increment, and to calculate the transient thermal impedance weighted thermal age centroid increment based on the mapping relationship.

[0011] The sensitivity calculation unit is used to apply a minimal perturbation to the transient thermal impedance weighted thermal age centroid and output the local temperature sensitivity in combination with the long short-term memory neural network.

[0012] The adjustment unit is used to solve for the minimum change in heat injection timing displacement and heat injection amplitude scaling based on the predicted temperature value of the zone, the local temperature sensitivity, and the weighted thermal age centroid increment of the transient thermal impedance, under the constraint that the predicted temperature value of the zone is not higher than the set temperature value of the zone. It generates a timing plan table and an amplitude plan table and sends them to the driver end of the LED transparent display screen.

[0013] The beneficial effects of this invention are as follows: by using the transient thermal impedance weighted thermal age centroid as the core quantity and combining it with the dynamic prediction capability of a long short-term memory neural network, it achieves proactive prediction and minimal adjustment of the temperature of multiple zones in a transparent LED display screen. This invention does not rely on traditional average power or real-time feedback, but rather establishes a mapping relationship between heat injection timing displacement, heat injection amplitude scaling, and temperature prediction results, accurately reflecting the effects of local heat accumulation and timing thermal delay. This invention can adjust the zone temperature before it approaches the set value, effectively suppressing thermal oscillations and local overheating caused by scanning phase changes, thereby significantly improving the stability and response speed of temperature control, delaying the thermal aging process of the display screen's light-emitting units, and maintaining overall brightness consistency and lifespan uniformity. Attached Figure Description

[0014] Figure 1 This is a flowchart of an anti-aging temperature control system for an LED transparent display screen according to the present invention. Detailed Implementation

[0015] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0016] like Figure 1 As shown, an anti-aging temperature control system for an LED transparent display screen includes:

[0017] The acquisition unit is used to acquire the measured temperature, control cycle, window duration, occurrence time of heat injection event and equivalent energy of each zone, and to form a pulse sequence based on the occurrence time of heat injection event and equivalent energy.

[0018] The thermal age centroid calculation unit is used to calculate the transient thermal impedance weighted thermal age centroid based on the pulse sequence and the transient thermal impedance function.

[0019] The temperature prediction unit is used to output zoned temperature prediction values ​​through a long short-term memory neural network based on the historical sequence of measured temperatures and the historical sequence of transient thermal impedance weighted thermal age centroids.

[0020] The mapping unit is used to establish the mapping relationship between the heat injection time-series displacement and the heat injection amplitude scaling to the transient thermal impedance weighted thermal age centroid increment, and to calculate the transient thermal impedance weighted thermal age centroid increment based on the mapping relationship.

[0021] The sensitivity calculation unit is used to apply a minimal perturbation to the transient thermal impedance weighted thermal age centroid and output the local temperature sensitivity in combination with the long short-term memory neural network.

[0022] The adjustment unit is used to solve for the minimum change in heat injection timing displacement and heat injection amplitude scaling based on the predicted temperature value of the zone, the local temperature sensitivity, and the weighted thermal age centroid increment of the transient thermal impedance, under the constraint that the predicted temperature value of the zone is not higher than the set temperature value of the zone. It generates a timing plan table and an amplitude plan table and sends them to the driver end of the LED transparent display screen.

[0023] In one embodiment of the present invention, the measured temperature, control cycle, window duration, occurrence time of heat injection event, and equivalent energy of each partition are collected, and a pulse sequence is formed based on the occurrence time of heat injection event and equivalent energy, including:

[0024] The initial time is determined, and the sampling time is calculated based on the control period; wherein, the sampling time is the sum of the initial time and an integer multiple of the control period;

[0025] The time window is set according to the sampling time and the window duration. The time window is the time interval formed by shifting the window duration forward from the sampling time.

[0026] Filter out the hot injection events that occur within the time window to form a set of hot injection events within the window;

[0027] A pulse sequence is constructed based on the occurrence time and equivalent energy of each heat injection event in the set of heat injection events within the window; wherein, the position of each pulse in the pulse sequence corresponds to the occurrence time of the heat injection event, and the energy value of each pulse corresponds to the equivalent energy of the heat injection event.

[0028] The initial time is the starting point when the system starts up and begins collecting data. It is usually set as the first time point after the system completes its self-test after power-on. It can be recorded using absolute time (such as a certain year, month, day, hour, minute, and second) or relative time (such as 0 seconds after system startup). It must be consistent with the time base of all subsequent time parameters (such as the sampling time and the time when the hot injection event occurs).

[0029] The control cycle is a fixed time interval between two consecutive data acquisitions and parameter calculations. It must match the frame cycle of the LED transparent screen or be an integer multiple of the frame cycle (e.g., a screen refresh rate of 1920 Hz corresponds to a frame cycle of approximately 0.52 milliseconds, and the control cycle can be set to 50 milliseconds, which is approximately 96 frame cycles) to ensure that heat injection events within multiple frames can be captured synchronously and to avoid timing misalignment.

[0030] The sampling time is the point in time when data is collected and subsequent calculations (such as thermal age centroid calculation) are triggered. It needs to be generated according to a fixed period. Specifically, the nth sampling time is equal to the initial time plus the control period multiplied by n, where n is a positive integer. n starts from 1 and increases sequentially, corresponding to the 1st, 2nd and so on, nth sampling time (for example, if the initial time is 0 seconds and the control period is 50 milliseconds, the 1st sampling time is 0 seconds + 50 milliseconds × 1 = 50 milliseconds, and the 2nd sampling time is 0 seconds + 50 milliseconds × 2 = 100 milliseconds).

[0031] The window duration is the length of time for reviewing historical heat injection events during each calculation. It must cover all historical heat injections that affect the current temperature (based on the LED thermal time constant). Specifically, the window duration is taken as 5 times the thermal time constant of the LED device used in the LED transparent screen. The thermal time constant is obtained from the device datasheet provided by the LED manufacturer (commonly ranging from tens to hundreds of milliseconds). If the datasheet does not provide it directly, it is measured using the JEDEC JESD51-14 standard test method (for example, if the thermal time constant is 40 milliseconds, the window duration is set to 200 milliseconds).

[0032] The time window is the time range for filtering historical hot injection events. Only hot injection events within this range will participate in subsequent calculations. Specifically, the time window corresponding to each sampling moment has a start time equal to the sampling moment minus the window duration and an end time equal to the sampling moment. The time window is a closed interval from the start time to the end time (for example, if the sampling moment is 100 milliseconds and the window duration is 200 milliseconds, the time window is a closed interval from 100 milliseconds - 200 milliseconds = -100 milliseconds to 100 milliseconds. In actual filtering, the effective time after system startup is taken, that is, from the initial moment).

[0033] The occurrence time of a heat injection event is the actual time when heat is injected into a certain section of the LED transparent screen during the scanning drive and PWM dimming process. During the time period when each subframe lights up the section, the start time of each PWM pulse is the occurrence time of a heat injection event, which needs to be read from the timing log of the screen driver controller. The time base is consistent with the initial time.

[0034] The equivalent energy of a heat injection event is the total electrical energy generated by a single heat injection event that can be converted into heat. It needs to be calculated in conjunction with the LED electrical parameters and drive timing. Specifically, the equivalent energy is equal to the LED forward voltage multiplied by the peak current corresponding to the heat injection event, then multiplied by the PWM duty cycle of the heat injection event, and finally multiplied by the duration of the heat injection event. The LED forward voltage is obtained from the LED device datasheet (usually 2-3.5 volts), the peak current is read from the screen drive parameters (related to the 1 / N scan ratio, where N is 4, 8, 16, etc., and the smaller the scan ratio, the larger the peak current), the PWM duty cycle is read from the drive controller (a decimal between 0 and 1), and the duration is the width of the PWM pulse (usually in the microsecond range, obtained from the timing log).

[0035] The set of heat injection events within a window is a collection of all heat injection events whose occurrence time falls within the time window corresponding to the current sampling time. Each element in the set contains the occurrence time and equivalent energy of the heat injection event. Specifically, the occurrence time of all heat injection events recorded by the system is checked one by one to determine whether the time is between the start and end time of the time window. If it is, the heat injection event (including the occurrence time and equivalent energy) is included in the set.

[0036] A pulse sequence is a sequence that reflects the time distribution and energy magnitude of heat injection within a window. It needs to be arranged in chronological order to reflect the temporal characteristics of heat injection. Specifically, firstly, the events in the heat injection event set within the window are sorted from earliest to latest according to their occurrence time. Then, a pulse is marked on the time axis corresponding to the occurrence time of each event. The height or amplitude of the pulse is set as the equivalent energy of the event, forming a complete pulse sequence.

[0037] In one embodiment of the present invention, the transient thermal impedance weighted thermal age centroid is calculated based on the pulse sequence and the transient thermal impedance function, including:

[0038] Extract the occurrence time and equivalent energy of each heat injection event from the heat injection event set within the window from the pulse sequence;

[0039] Calculate the time interval between the occurrence time of each heat injection event and the sampling time;

[0040] Input each time interval into the transient thermal impedance function to obtain the thermal weight corresponding to each heat injection event;

[0041] The sum of the products of each heat weight, the equivalent energy of the heat injection event corresponding to that heat weight, and the time interval of the heat injection event corresponding to that heat weight is used as the weighted cumulative amount.

[0042] The sum of the products of each heat weight and the equivalent energy of the heat injection event corresponding to that heat weight is calculated as the normalized cumulative amount.

[0043] Divide the weighted cumulative value by the normalized cumulative value to obtain the weighted thermal age centroid of transient thermal impedance.

[0044] The time interval is the duration from the occurrence of the heat injection event to the current sampling time, reflecting the aging degree of the heat injection. The longer the duration, the smaller the impact of the event on the current temperature. Specifically, the time interval is equal to the sampling time minus the occurrence time of the heat injection event, and the result must be a non-negative value (if a negative value appears, it means that the event has exceeded the time window and should be removed).

[0045] The transient thermal impedance function is a function that describes the thermal conduction characteristics of LED devices. The output value changes with the time interval (the shorter the time interval, the larger the function value, which represents the stronger thermal conduction capability and the more significant the thermal effect). It needs to be obtained from the LED device datasheet or the JEDEC JESD51-14 standard test report (you can directly look up the transient thermal impedance curve of the device or calculate it through the RC thermal model of the device). Specifically, based on the calculated time interval, find the corresponding horizontal axis (time interval) vertical axis value on the transient thermal impedance curve, which is the function output value corresponding to that time interval.

[0046] Thermal weight is a coefficient that measures the contribution of a heat injection event to the current temperature. It is directly related to the output value of the transient thermal impedance function. The larger the weight, the more significant the impact of the event on the current temperature. Specifically, thermal weight is equal to the output value obtained by substituting the time interval of the heat injection event into the transient thermal impedance function.

[0047] The weighted cumulative amount is the cumulative value of the combined heat weight, the amount of heat injection energy, and the heat injection time interval. It reflects the total contribution of all heat injection events in the dimensions of heat impact degree, energy, and time. Specifically, for each heat injection event, the heat weight of the event is first calculated and multiplied by its equivalent energy, and then multiplied by its time interval to obtain the product term of a single event. Then, the product terms of all events are added together, and the result is the weighted cumulative amount.

[0048] The normalized cumulative amount is the cumulative value of the combined heat weight and the amount of heat injection energy. It is used to normalize the weighted cumulative amount and eliminate the influence of the difference in total energy on the calculation of the centroid of thermal age. Specifically, for each heat injection event, the heat weight of the event is calculated and multiplied by its equivalent energy to obtain the product term of a single event. Then, the product terms of all events are added together, and the result is the normalized cumulative amount.

[0049] The transient thermal impedance weighted thermal age centroid is a single scalar that reflects the equivalent thermal delay of all heat injection events within the window. The smaller the value, the closer the heat injection that plays a dominant role in the current temperature is to the current moment, and the larger the value, the earlier the dominant heat injection. Specifically, the transient thermal impedance weighted thermal age centroid is equal to the weighted cumulative amount divided by the normalized cumulative amount (if the normalized cumulative amount is zero, it means that there is no heat injection event within the window, and the thermal age centroid is set to half the duration of the time window).

[0050] In one embodiment of the present invention, based on the historical sequence of measured temperatures and the historical sequence of transient thermal impedance weighted thermal age centroids, a long short-term memory neural network is used to output predicted values ​​for zoned temperatures, including:

[0051] Select multiple consecutive sampling times, extract the measured temperature corresponding to each sampling time, and form a historical sequence of measured temperature;

[0052] Extract the transient thermal impedance weighted thermal age centroids corresponding to multiple consecutive sampling times to form a historical sequence of transient thermal impedance weighted thermal age centroids;

[0053] The historical sequence of measured temperature is combined with the historical sequence of transient thermal impedance weighted thermal age centroid to form the input sequence of the long short-term memory neural network;

[0054] The input sequence is fed into a trained long short-term memory neural network to obtain the predicted temperature value of the partition at the next sampling time.

[0055] Multiple consecutive sampling moments are sampling points within a time window used to construct the historical sequence. They need to cover the memory period (i.e., thermal time constant) of the LED thermal system to ensure that the historical data can reflect the impact of heat accumulation on the current temperature. Specifically, first obtain the thermal time constant of the LED device (obtained from the device datasheet or JEDEC test report), then divide the thermal time constant by the control period to obtain the base number, and actually select this base number to twice the base number of consecutive sampling moments (for example, if the thermal time constant is 50 milliseconds, the control period is 10 milliseconds, the base number is 5, select 5 to 10 consecutive sampling moments).

[0056] The historical sequence of measured temperatures is a set of measured temperatures from multiple consecutive sampling times arranged in chronological order, used to reflect the temporal trend of temperature change. Specifically, the selected multiple consecutive sampling times are sorted from early to late, and the measured temperature corresponding to each time time is extracted in turn, forming a one-dimensional sequence according to the sorting order (for example, if times t1, t2, and t3 are selected, the corresponding temperatures are T1, T2, and T3, and the sequence is [T1, T2, T3]).

[0057] The historical sequence of transient thermal impedance weighted thermal age centroids is a set of thermal age centroids arranged in chronological order for multiple consecutive sampling times, used to reflect the changing trend of the thermal injection time sequence. Specifically, the selected multiple consecutive sampling times are sorted from early to late, and the thermal age centroids corresponding to each time time are extracted in turn, forming a one-dimensional sequence according to the sorting order (for example, selecting times t1, t2, t3, corresponding to thermal age centroids H1, H2, H3, the sequence is [H1, H2, H3]).

[0058] The input sequence is two-dimensional data that integrates temperature time series and heat injection time series information. It is used to provide the historical features required for neural network prediction. Specifically, it is based on time steps. Each time step corresponds to a sampling moment. The historical sequence elements of the measured temperature at that moment are concatenated with the historical sequence elements of the thermal age centroid (i.e., each time step contains two features: temperature and thermal age centroid). The sequence is arranged from early to late according to the time steps to form an input sequence with a shape of time step number × 2 (for example, for 5 time steps, the input sequence is [[T1,H1],[T2,H2],[T3,H3],[T4,H4],[T5,H5]]).

[0059] The trained Long Short-Term Memory (LSTM) neural network is a model capable of temperature prediction. It needs to be trained to convergence under supervision using historical data. Specifically, the first step is to collect historical data (including temperature sequences, thermal age centroid sequences, and the measured temperature at the next corresponding moment) from multiple consecutive sampling times. The second step is to divide the data into a training set (70% of the total data), a validation set (20%), and a test set (10%). The third step is to set the loss function as mean squared error (which measures the deviation between the predicted and measured temperatures). The fourth step is to use the Adam optimizer with a learning rate of 0.001. After each training round, the validation set loss is calculated. Training stops when the validation set loss does not decrease for five consecutive rounds, resulting in a fully trained model.

[0060] The zone temperature prediction is the zone temperature estimate at the next sampling time, output by the neural network based on the historical sequence of the input.

[0061] In one embodiment of the present invention, a mapping relationship is established between the heat injection time-series displacement and the heat injection amplitude scaling to the transient thermal impedance weighted thermal age centroid increment, and the transient thermal impedance weighted thermal age centroid increment is calculated using the mapping relationship, including:

[0062] From the set of heat injection events within the window, heat injection events with a time interval not exceeding the control cycle are selected to form a near-window heat injection event set;

[0063] Apply a small advance amount to the occurrence time of each heat injection event in the near-window heat injection event set, calculate the difference in the centroid of the transient thermal impedance weighted thermal age before and after the application, and use the ratio of this difference to the small advance amount as the time displacement coefficient.

[0064] Apply a small amplification factor to the equivalent energy of each heat injection event in the near-window heat injection event set, calculate the difference in the centroid of transient thermal impedance weighted thermal age before and after the amplification factor, and use the ratio of the difference in the centroid of transient thermal impedance weighted thermal age to the small amplification factor as the amplitude scaling factor.

[0065] The mapping relationship is obtained by adding the product of the timing displacement coefficient and the heat injection timing displacement and the control period, and the product of the amplitude scaling coefficient and the heat injection amplitude scaling minus one.

[0066] Based on the mapping relationship, by substituting the heat injection time-series displacement and heat injection amplitude scaling, the transient thermal impedance weighted thermal age centroid increment is calculated.

[0067] The near-window heat injection event set is a subset of heat injection events within the window that are most sensitive to temperature changes in the next cycle, and only includes events that have occurred recently. Specifically, the time interval of each event in the heat injection event set within the window is checked one by one to determine whether the time interval is less than or equal to the control cycle. If it is, the event (including the time of occurrence and equivalent energy) is included in the near-window heat injection event set.

[0068] The small advance is a virtual advance duration applied to the occurrence time of near-window events for the purpose of calculating the timing displacement coefficient. It must be much smaller than the control period (to avoid changing the near-window attribute of the event). Specifically, the small advance is 0.1 to 0.5 times the control period (for example, if the control period is 50 milliseconds, the small advance is 5 to 25 milliseconds). The specific value must ensure that the occurrence time of the event is still within the time window after the advance is applied.

[0069] The thermal age centroid before application is the thermal age centroid at the current sampling time; the thermal age centroid after application is the thermal age centroid recalculated after subtracting a small advance amount from the occurrence time of the near-window event.

[0070] The timing displacement coefficient is a coefficient that measures the degree of influence of advancing the heat injection timing on the thermal age centroid. A negative value means that advancing the timing will reduce the thermal age centroid (which is logical: the earlier the event occurs, the larger the thermal age, and the earlier the event occurs, the smaller the thermal age). Specifically, the timing displacement coefficient is equal to (the thermal age centroid after applying a small advance minus the thermal age centroid before application) divided by the small advance.

[0071] The micro-magnification factor is a virtual amplification ratio applied to the equivalent energy of a near-window event for calculating the amplitude scaling factor. It needs to be slightly greater than 1 (to avoid excessive energy change). Specifically, the micro-magnification factor is between 1.05 and 1.2 (e.g., 1.1, which represents 10% energy amplification). The specific value must ensure that the energy remains within the safe operating range of the LED device after application.

[0072] The amplitude scaling factor is a coefficient that measures the degree of influence of thermal injection energy amplification on the thermal age centroid. The value is usually positive (after energy amplification, the proportion of thermal contribution of recent events increases, the thermal age centroid decreases, and the difference is negative, which needs to be judged in combination with the actual sign); specifically, the amplitude scaling factor is equal to (the thermal age centroid after applying the small amplification factor minus the thermal age centroid before application) divided by (the small amplification factor minus 1).

[0073] The heat injection timing shift is the proportion (dimensionless) of the heat injection event in the next cycle relative to the original planned time, reflecting the manipulation amount of adjusting the heat injection occurrence time; specifically, the heat injection timing shift is taken from 0 to 1 (for example, 0.2 means advancing the heat injection event in the next cycle by 0.2 times the control cycle), 0 means not advancing, 1 means advancing by the entire control cycle, and cannot be negative (to avoid the event exceeding the time window).

[0074] Heat injection amplitude scaling is the scaling ratio (dimensionless) of the equivalent energy of the heat injection event in the next cycle, reflecting the amount of manipulation to adjust the size of the heat injection energy; specifically, heat injection amplitude scaling is between 0.5 and 1 (for example, 0.8 means reducing the heat injection energy in the next cycle to 80% of the original energy), 1 means no scaling, and the lower limit of 0.5 is to avoid the brightness being too low and affecting the display effect.

[0075] The mapping relationship is an expression that quantifies the linear relationship between the thermal injection time-series displacement, amplitude scaling and thermal age centroid increment, and is used to convert the manipulated quantity into the change of thermal age centroid; specifically, the mapping relationship is equal to (time-series displacement coefficient multiplied by thermal injection time-series displacement multiplied by control period) plus (amplitude scaling coefficient multiplied by (thermal injection amplitude scaling minus 1)).

[0076] The thermal age centroid increment is the change in the thermal age centroid relative to the current moment after the thermal injection parameters are adjusted in the next cycle. A negative value indicates that the thermal age centroid decreases (the thermal injection is closer to the current moment). Specifically, the transient thermal impedance weighted thermal age centroid increment is equal to the result obtained by substituting the thermal injection time-series displacement and thermal injection amplitude scaling into the mapping relationship.

[0077] In one embodiment of the present invention, a minimal perturbation is applied to the transient thermal impedance weighted thermal age centroid, and the local temperature sensitivity is output in conjunction with a long short-term memory neural network, including:

[0078] Select the last transient thermal impedance weighted thermal age centroid in the historical sequence of transient thermal impedance weighted thermal age centroids, apply a minimal perturbation to this transient thermal impedance weighted thermal age centroid, and obtain the perturbed transient thermal impedance weighted thermal age centroid;

[0079] Replace the last transient thermal impedance weighted thermal age centroid in the historical sequence of transient thermal impedance weighted thermal age centroids with the perturbed transient thermal impedance weighted thermal age centroids to form the historical sequence of perturbed transient thermal impedance weighted thermal age centroids.

[0080] The historical sequence of measured temperature is combined with the historical sequence of the perturbed transient thermal impedance weighted thermal age centroid to form the input sequence of the perturbed long short-term memory neural network.

[0081] The perturbated input sequence is fed into a trained long short-term memory neural network to obtain the predicted temperature of the partition at the next sampling time after perturbation.

[0082] Calculate the difference between the predicted temperature of the zone after the perturbation and the predicted temperature of the zone without the perturbation, and divide the difference by the amplitude of the minimum perturbation to obtain the local temperature sensitivity.

[0083] The last transient thermal impedance weighted thermal age centroid is the latest element in the thermal age centroid history sequence (corresponding to the thermal age centroid at the most recent sampling moment). Its influence on the prediction of the next temperature moment by the long short-term memory neural network is the most significant. Therefore, this element is selected to apply perturbation to accurately capture the marginal influence of the current thermal time series on the temperature.

[0084] The minimum perturbation is a tiny increment applied to the last thermal age centroid for calculating local temperature sensitivity. It must be small and reasonable (without changing the temporal properties of the thermal age centroid). Specifically, the amplitude of the minimum perturbation is 0.01 to 0.05 times the value of the last transient thermal impedance-weighted thermal age centroid, and the maximum value does not exceed 10 milliseconds (for example, if the last thermal age centroid is 50 milliseconds, the minimum perturbation is 0.5 to 2.5 milliseconds), to ensure that the thermal age centroid can still reflect the true thermal temporal characteristics after the perturbation.

[0085] The perturbation-weighted transient thermal impedance centroid of thermal age is the result of superimposing the last thermal age centroid with a minimal perturbation, used to simulate scenarios with small changes in the thermal age centroid; specifically, the perturbation-weighted transient thermal impedance centroid of thermal age is equal to the last transient thermal impedance centroid of thermal age plus a minimal perturbation (only positive perturbations are used to avoid negative thermal age centroids being less than zero).

[0086] The perturbated thermal age centroid history sequence is a sequence that only replaces the latest thermal age centroid in the original sequence, while retaining other historical elements to ensure the continuity of the time series trend. Specifically, all elements in the transient thermal impedance weighted thermal age centroid history sequence except the last element are retained, and the last element of the original sequence is replaced with the perturbated transient thermal impedance weighted thermal age centroid, and the new sequence is formed by arranging them in the original time order.

[0087] The perturbated input sequence is a two-dimensional data fusion of the unchanged temperature history and the perturbated thermal age centroid history, with the same format as the input sequence. Specifically, it is based on time steps, with each time step corresponding to a sampling moment. The elements of the measured temperature history sequence at that moment are concatenated with the elements of the perturbated thermal age centroid history sequence (each time step contains two features: temperature and perturbated thermal age centroid), and arranged from early to late according to the time steps to form a sequence.

[0088] The perturbed zone temperature prediction is the temperature estimate at the next sampling time output by the trained neural network on the perturbed input sequence, used to compare with the unperturbed prediction to calculate sensitivity.

[0089] The temperature prediction value of the partition without perturbation is the temperature prediction value at the next sampling moment of the trained neural network output by inputting the unperturbed input sequence (original temperature history + original thermal age centroid history), which is the benchmark value for sensitivity calculation.

[0090] Local temperature sensitivity is a coefficient that measures the degree of influence of small changes in the thermal age centroid on the temperature prediction value at the next moment. A positive value means that the temperature prediction value increases when the thermal age centroid increases (logically, the larger the thermal age, the earlier the heat injection, and the lower the current temperature, so it is usually a negative value, and needs to be judged in conjunction with the actual sign). Specifically, local temperature sensitivity is equal to (the temperature prediction value of the zone after the perturbation minus the temperature prediction value of the zone without the perturbation) divided by the amplitude of the smallest perturbation.

[0091] In one embodiment of the present invention, based on the predicted zone temperature, local temperature sensitivity, and transient thermal impedance weighted thermal age centroid increment, and under the constraint that the predicted zone temperature does not exceed the zone temperature setpoint, the minimum modified heat injection timing shift and heat injection amplitude scaling are solved, a timing schedule table and an amplitude schedule table are generated, and sent to the driver of the LED transparent display screen, including:

[0092] The difference between the zone temperature setpoint and the zone temperature prediction value is calculated and used as the upper limit of the constraint.

[0093] Construct an objective function that is the product of the square of the heat injection time displacement and the first weight, plus the product of the square of the difference between the heat injection amplitude scaling and one and the second weight.

[0094] Based on the mapping relationship between local temperature sensitivity, transient thermal impedance weighted thermal age centroid increment and constraint upper limit, the objective function is solved to obtain the heat injection time-series displacement and heat injection amplitude scaling.

[0095] Apply upper and lower limits to the obtained heat injection timing displacement and apply upper and lower limits to the obtained heat injection amplitude scaling; fill the limited heat injection timing displacement into the timing plan table and fill the limited heat injection amplitude scaling into the amplitude plan table.

[0096] The timing schedule and amplitude schedule are sent to the driver of the LED transparent display screen.

[0097] The zone temperature setting is the highest safe temperature allowed for a zone by the system. It should be set according to the thermal safety threshold in the LED device manual (e.g., the safe upper limit of the common LED junction temperature is 60 degrees Celsius, and the zone surface temperature of the screen is set to 50 degrees Celsius) to avoid excessive temperature from causing accelerated aging or failure of the LED.

[0098] The upper limit of the constraint is the available space for temperature adjustment, reflecting the maximum allowable increase in temperature in the next cycle. If the difference is negative, it means that the current predicted temperature has exceeded the set value and forced adjustment is required. Specifically, the upper limit of the constraint is equal to the zone temperature set value minus the zone temperature predicted value. If the calculation result is less than 0, the upper limit of the constraint is 0 (ensuring that the temperature must be pulled back below the set value through adjustment).

[0099] The objective function is an indicator that measures the magnitude of changes in the heat injection parameters. The smaller the value, the smaller the change (and the smaller the impact on the display effect). The core is to balance the adjustment cost of timing shift and amplitude scaling. Specifically, the objective function is equal to (the first weight multiplied by the square of the heat injection timing shift) plus (the second weight multiplied by the square of (heat injection amplitude scaling minus 1)).

[0100] The first weight is a coefficient that balances the priority of heat injection timing shift changes. It ranges from 0 to 1. When timing adjustment is prioritized (such as avoiding brightness reduction), a smaller value is used, and when temperature adjustment is prioritized, a larger value is used. Specifically, if it is necessary to prioritize ensuring display brightness (reducing amplitude scaling), the first weight is 0.2 to 0.4; if it is necessary to control the temperature quickly (allowing for larger timing changes), the first weight is 0.6 to 0.8.

[0101] The second weight is a coefficient that balances the priority of heat injection amplitude scaling changes. It ranges from 0 to 1 and is equal to 1 minus the first weight. The larger value is taken when amplitude adjustment takes priority (such as when timing changes are ineffective). Specifically, the second weight is equal to 1 minus the first weight. If the first weight is 0.3, the second weight is 0.7.

[0102] The solution process involves finding the combination of control quantities that minimizes the objective function while keeping the temperature within the setpoint constraint, using the Lagrange multiplier method. Specifically, the first step is to transform the temperature constraint (the predicted temperature value of the zone plus the local temperature sensitivity multiplied by the weighted centroid increment of the transient thermal impedance, the result of which is less than or equal to the setpoint temperature of the zone) into an equality constraint (taking the equality sign, i.e., maximizing the utilization of the adjustment space). The second step is to construct the Lagrange function (the objective function plus the Lagrange multiplier multiplied by the constraint equation). The third step is to calculate the partial derivatives of the Lagrange function with respect to the heat injection time-series displacement and the heat injection amplitude scaling, set the partial derivatives to 0, and solve the constraint equations simultaneously to obtain the two control quantities.

[0103] The upper and lower limits are the boundaries that ensure the timing offset conforms to logic (to avoid hot injection events exceeding the time window or being introduced prematurely for no reason). Specifically, the upper limit of the hot injection timing offset is 1, and the lower limit is 0. If the solution result is greater than 1, it is set to 1; if it is less than 0, it is set to 0.

[0104] The upper and lower limits are the boundaries to ensure that the amplitude scaling meets the display requirements (to avoid the brightness being too low and affecting the viewing experience); specifically, the upper limit of the heat injection amplitude scaling is 1, and the lower limit is 0.5. If the solution result is greater than 1, then 1 is used, and if it is less than 0.5, then 0.5 is used.

[0105] The timing schedule table is a table that records the timing adjustment parameters for the next cycle of hot injection. It only contains the limited hot injection timing displacement, and its format matches the timing control protocol of the screen driver.

[0106] The amplitude plan table is a table that records the adjustment parameters of the heat injection energy in the next cycle. It contains only the limited heat injection amplitude scaling (single value) and the format matches the PWM / current control protocol of the screen driver.

[0107] The driver of an LED transparent display screen is a hardware unit (such as a scanning driver chip or PWM controller) that receives and executes adjustment parameters. It can adjust the position of the partition points according to the timing schedule and adjust the PWM duty cycle or peak current according to the amplitude schedule.

[0108] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. An anti-aging temperature control system for an LED transparent display screen, characterized in that, include: The acquisition unit is used to acquire the measured temperature, control cycle, window duration, occurrence time of heat injection event and equivalent energy of each zone, and to form a pulse sequence based on the occurrence time of heat injection event and equivalent energy. The thermal age centroid calculation unit is used to calculate the transient thermal impedance weighted thermal age centroid based on the pulse sequence and the transient thermal impedance function, including: Extract the occurrence time and equivalent energy of each heat injection event from the heat injection event set within the window from the pulse sequence; Calculate the time interval between the occurrence time of each heat injection event and the sampling time; Input each time interval into the transient thermal impedance function to obtain the thermal weight corresponding to each heat injection event; The sum of the products of each heat weight, the equivalent energy of the heat injection event corresponding to that heat weight, and the time interval of the heat injection event corresponding to that heat weight is used as the weighted cumulative amount. The sum of the products of each heat weight and the equivalent energy of the heat injection event corresponding to that heat weight is calculated as the normalized cumulative amount. Divide the weighted cumulative amount by the normalized cumulative amount to obtain the weighted thermal age centroid of transient thermal impedance; The temperature prediction unit is used to output zoned temperature prediction values ​​through a long short-term memory neural network based on the historical sequence of measured temperatures and the historical sequence of transient thermal impedance weighted thermal age centroids. The mapping unit is used to establish the mapping relationship between the heat injection time-series displacement and the heat injection amplitude scaling to the transient thermal impedance weighted thermal age centroid increment, and to calculate the transient thermal impedance weighted thermal age centroid increment based on the mapping relationship. The sensitivity calculation unit is used to apply a minimal perturbation to the transient thermal impedance weighted thermal age centroid and output the local temperature sensitivity in combination with the long short-term memory neural network. The adjustment unit is used to solve for the minimum change in heat injection timing displacement and heat injection amplitude scaling based on the predicted temperature value of the zone, the local temperature sensitivity, and the weighted thermal age centroid increment of the transient thermal impedance, under the constraint that the predicted temperature value of the zone is not higher than the set temperature value of the zone. It generates a timing plan table and an amplitude plan table and sends them to the driver end of the LED transparent display screen.

2. The anti-aging temperature control system for an LED transparent display screen according to claim 1, characterized in that, The system collects measured temperatures, control cycles, window durations, occurrence times of heat injection events, and equivalent energy for each zone. Based on the occurrence times and equivalent energy of the heat injection events, a pulse sequence is generated, including: The initial time is determined, and the sampling time is calculated based on the control period; wherein, the sampling time is the sum of the initial time and an integer multiple of the control period; The time window is set according to the sampling time and the window duration. The time window is the time interval formed by shifting the window duration forward from the sampling time. Filter out the hot injection events that occur within the time window to form a set of hot injection events within the window; A pulse sequence is constructed based on the occurrence time and equivalent energy of each heat injection event in the set of heat injection events within the window; wherein, the position of each pulse in the pulse sequence corresponds to the occurrence time of the heat injection event, and the energy value of each pulse corresponds to the equivalent energy of the heat injection event.

3. The anti-aging temperature control system for an LED transparent display screen according to claim 2, characterized in that, Based on the historical sequence of measured temperatures and the historical sequence of transient thermal impedance weighted thermal age centroids, a long short-term memory neural network is used to output predicted regional temperatures, including: Select multiple consecutive sampling times, extract the measured temperature corresponding to each sampling time, and form a historical sequence of measured temperature; Extract the transient thermal impedance weighted thermal age centroids corresponding to multiple consecutive sampling times to form a historical sequence of transient thermal impedance weighted thermal age centroids; The historical sequence of measured temperature is combined with the historical sequence of transient thermal impedance weighted thermal age centroid to form the input sequence of the long short-term memory neural network; The input sequence is fed into a trained long short-term memory neural network to obtain the predicted temperature value of the partition at the next sampling time.

4. The anti-aging temperature control system for an LED transparent display screen according to claim 3, characterized in that, Establish a mapping relationship between the heat injection time-series displacement and the heat injection amplitude scaling to the transient thermal impedance weighted thermal age centroid increment, and calculate the transient thermal impedance weighted thermal age centroid increment using this mapping relationship, including: From the set of heat injection events within the window, heat injection events with a time interval not exceeding the control cycle are selected to form a near-window heat injection event set; Apply a small advance amount to the occurrence time of each heat injection event in the near-window heat injection event set, calculate the difference in the centroid of the transient thermal impedance weighted thermal age before and after the application, and use the ratio of this difference to the small advance amount as the time displacement coefficient. Apply a small amplification factor to the equivalent energy of each heat injection event in the near-window heat injection event set, calculate the difference in the centroid of transient thermal impedance weighted thermal age before and after the amplification factor, and use the ratio of the difference in the centroid of transient thermal impedance weighted thermal age to the small amplification factor as the amplitude scaling factor. The mapping relationship is obtained by adding the product of the timing displacement coefficient and the heat injection timing displacement and the control period, and the product of the amplitude scaling coefficient and the heat injection amplitude scaling minus one. Based on the mapping relationship, by substituting the heat injection time-series displacement and heat injection amplitude scaling, the transient thermal impedance weighted thermal age centroid increment is calculated.

5. The anti-aging temperature control system for an LED transparent display screen according to claim 4, characterized in that, Applying a minimal perturbation to the transient thermal impedance-weighted thermal age centroid, and combining this with a long short-term memory neural network to output local temperature sensitivity, including: Select the last transient thermal impedance weighted thermal age centroid in the historical sequence of transient thermal impedance weighted thermal age centroids, apply a minimal perturbation to this transient thermal impedance weighted thermal age centroid, and obtain the perturbed transient thermal impedance weighted thermal age centroid; Replace the last transient thermal impedance weighted thermal age centroid in the historical sequence of transient thermal impedance weighted thermal age centroids with the perturbed transient thermal impedance weighted thermal age centroids to form the historical sequence of perturbed transient thermal impedance weighted thermal age centroids. The historical sequence of measured temperature is combined with the historical sequence of the perturbed transient thermal impedance weighted thermal age centroid to form the input sequence of the perturbed long short-term memory neural network. The perturbated input sequence is fed into a trained long short-term memory neural network to obtain the predicted temperature of the partition at the next sampling time after perturbation. Calculate the difference between the predicted temperature of the zone after the perturbation and the predicted temperature of the zone without the perturbation, and divide the difference by the amplitude of the minimum perturbation to obtain the local temperature sensitivity.

6. The anti-aging temperature control system for an LED transparent display screen according to claim 5, characterized in that, Based on the predicted zone temperature, local temperature sensitivity, and weighted thermal age centroid increment of transient thermal impedance, and under the constraint that the predicted zone temperature does not exceed the zone temperature setpoint, the minimum altered heat injection timing shift and heat injection amplitude scaling are calculated. A timing schedule and amplitude schedule are then generated and sent to the driver of the LED transparent display screen, including: The difference between the zone temperature setpoint and the zone temperature prediction value is calculated and used as the upper limit of the constraint. Construct an objective function that is the product of the square of the heat injection time displacement and the first weight, plus the product of the square of the difference between the heat injection amplitude scaling and one and the second weight. Based on the mapping relationship between local temperature sensitivity, transient thermal impedance weighted thermal age centroid increment and constraint upper limit, the objective function is solved to obtain the heat injection time-series displacement and heat injection amplitude scaling. Apply upper and lower limits to the obtained heat injection timing displacement and apply upper and lower limits to the obtained heat injection amplitude scaling; fill the limited heat injection timing displacement into the timing plan table and fill the limited heat injection amplitude scaling into the amplitude plan table. The timing schedule and amplitude schedule are sent to the driver of the LED transparent display screen.

Images