Miniature LED display intelligent regulation and control system

Abstract

The invention relates to the technical field of LED control and adjustment, in particular to a miniature LED display intelligent regulation and control system which comprises a driving control unit, a state prediction unit, a dynamic compensation unit and a self-learning calibration unit. The driving control unit provides a row gating duration, an initial voltage amplitude and an auxiliary reference capacitance reference attenuation rate; the state prediction unit collects the residual voltage of the storage capacitor, the regional temperature and the total working time of pixels through a multi-dimensional monitoring module, calculates an attenuation rate difference value in combination with reference data, and predicts an attenuation trend. The dynamic compensation unit adjusts an initial voltage and a gating time sequence through feedforward compensation, carries out feedback compensation and real-time monitoring, injects a compensation voltage to offset attenuation, forms a closed-loop control link through layered cooperation, actively deals with environment temperature changes and device aging, maintains the matching between TFT conduction time and a voltage maintenance period, and improves the performance of the device. And the brightness uniformity of the same gray scale area of the display picture is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of LED control and regulation, and particularly to an intelligent control system for micro-LED displays. Background Art

[0002] An LED display matrix is a display structure composed of pixel units arranged in a row-column rule. Each pixel unit includes an LED diode and a thin-film transistor (TFT) driving circuit. Among them, the TFT driving circuit integrates row scanning lines, column data lines, and storage capacitors. The row scanning lines selectively activate pixel units row by row through timing control signals. The column data lines transmit voltage signals corresponding to the image grayscale to the storage capacitors of the selectively activated pixel units. The storage capacitors maintain the voltage signals to enable the TFTs to enter the conducting state, thereby adjusting the magnitude of the flowing current and achieving dynamic control of the emission brightness.

[0003] The core technical pain point of intelligent control of an LED display matrix lies in the coupling control problem between the voltage maintenance stability of the storage capacitor and the consistency of dynamic brightness regulation. As a temporary storage unit for voltage signals in the TFT driving circuit, the voltage maintenance ability of the storage capacitor directly determines the duration of the conducting state of the TFT and the current regulation accuracy. When the environmental temperature changes or the dielectric material performance of the storage capacitor drifts during long-term device operation, the duration of voltage signal maintenance by the storage capacitor will fluctuate, which will cause the duration of the conducting state of the TFT to deviate from the design threshold, and further cause the magnitude of the current flowing through the LED diode to be unable to stably correspond to the image grayscale signal, ultimately resulting in uneven brightness in the same grayscale area of the display screen. For example, in a high-temperature environment, the leakage current of the storage capacitor increases, and the attenuation rate of the stored voltage signal accelerates. The TFT enters the cut-off state in advance, resulting in a shortened emission time of the LED diode corresponding to the pixel and an actual brightness lower than the target brightness corresponding to the grayscale signal, causing a brightness deviation in the local display area. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an intelligent control system for micro-LED displays to solve the problem of uneven brightness in the same grayscale area of the display screen caused by the failure of the coupling control between the voltage maintenance stability of the storage capacitor and the consistency of dynamic brightness regulation.

[0005] To solve the above technical problems, the specific technical solutions of the present invention are as follows: The micro-LED display intelligent control system provided by the present invention includes a driving control unit, a state prediction unit, a dynamic compensation unit, and a self-learning calibration unit; the driving control unit outputs the row strobe duration (high-level duration) through the row scanning sub-circuit, outputs the initial voltage amplitude of the storage capacitor through the column data sub-circuit, and outputs the real-time attenuation rate to the state prediction unit through the auxiliary reference capacitor; the state prediction unit includes a multi-dimensional monitoring module (composed of a voltage sampling sub-module, a temperature sensing sub-module, and a usage duration recording sub-module), which collects the residual voltage of the storage capacitor through the voltage sampling sub-module and calculates the real-time attenuation rate, collects the average regional temperature through the temperature sensing sub-module, accumulates the total pixel working duration through the usage duration recording sub-module, combines the initial voltage, row strobe duration, and auxiliary reference capacitor attenuation rate output by the driving control unit, and uses the historical database and pattern matching algorithm in the trend prediction module to predict the attenuation trend of the storage capacitor, and transmits the prediction result to the dynamic compensation unit; the dynamic compensation unit adjusts the voltage amplitude of the column data sub-circuit and the strobe timing of the row scanning sub-circuit through the feed-forward compensation module, and monitors the real-time attenuation rate of the storage capacitor during pixel emission through the feedback compensation module and feeds it back to the trend prediction module of the state prediction unit for model verification; the self-learning calibration unit generates a "decay rate and brightness deviation" mapping table by calling the redundant pixels at the edge of the display matrix through the hardware calibration module, and the model calibration module updates the prediction parameters of the historical database of the state prediction unit based on the decay rate and brightness deviation mapping table, and synchronously corrects the compensation parameters of the feed-forward module and the feedback module of the dynamic compensation unit to form a closed-loop control link of "driving data providing - state prediction - dynamic compensation - calibration optimization", so as to solve the problem of uneven brightness in the same gray-scale area of the display screen caused by the coupling failure of the voltage maintenance stability of the storage capacitor and the dynamic brightness regulation consistency.

[0006] Furthermore, in the micro-LED display intelligent control system of the present invention, the drive control unit includes a row scanning sub-circuit, a column data sub-circuit, and a dual-capacitance module; the row scanning sub-circuit generates a row selection signal (Gate signal) by a timing controller, and outputs a high-level signal row by row in the order of row addresses (such as the 1st row → the 2nd row →... → the nth row), activating the TFT gate of the corresponding row pixel unit to enter the conduction state; the column data sub-circuit generates a column data signal (Data signal) by a data driver, and writes an initial voltage signal corresponding to the image gray level (such as 5V for 100% gray level and 2.5V for 50% gray level) to the storage capacitors of all pixel units in this row only during the period when the corresponding row selection signal is at a high level (i.e., the TFT conduction stage); the dual-capacitance module includes a storage capacitor connected in parallel between the TFT gate and the ground (for temporarily storing the voltage signal to maintain the TFT conduction) and an auxiliary reference capacitor integrated in the non-display area of the display matrix (prepared with the same dielectric material and electrode process as the storage capacitor, for providing the reference attenuation rate data without display load interference); the row scanning sub-circuit outputs the row selection duration (the duration of the high level) to the voltage sampling sub-module of the state prediction unit, the column data sub-circuit outputs the initial voltage amplitude of the storage capacitor to the voltage sampling sub-module of the state prediction unit, and the auxiliary reference capacitor outputs the real-time attenuation rate to the trend prediction module of the state prediction unit through a micro-current monitoring circuit.

[0007] Furthermore, in the micro-LED display intelligent control system of the present invention, the state prediction unit includes a multi-dimensional monitoring module; the multi-dimensional monitoring module includes a voltage sampling sub-module, a temperature sensing sub-module, and a usage duration recording sub-module; the voltage sampling sub-module, at the end of each row strobe period (i.e., the instant when the row strobe signal jumps from high level to low level), collects the residual voltage value of the storage capacitor through a micro-current sampling circuit connected in parallel across the two ends of the storage capacitor, and calculates the attenuation rate of the storage capacitor within the real-time period based on the initial voltage amplitude output by the column data sub-circuit of the drive control unit and the row strobe duration output by the row scanning sub-circuit (attenuation rate = (initial voltage amplitude - residual voltage value) / row strobe duration); the temperature sensing sub-module integrates thin-film thermocouples in the pixel gap area of the TFT substrate (the center position of each 2×2 pixel unit), and collects the temperature values of each distribution point in real time (such as the temperature T1 of point 1 and the temperature T2 of point 2). After converting the analog temperature signal into a digital signal through an analog-to-digital conversion circuit, it calculates the average temperature of all distribution points (T_avg = (T1 + T2 +... + Tn) / n) as the average regional temperature; the usage duration recording sub-module configures a non-volatile counter (such as an OTP memory) for each pixel unit, and only accumulates and records the conduction time during the period when the TFT is conducting (i.e., the period when the row strobe signal is at high level). When the system starts, it reads and updates the total working duration of the pixel unit; the multi-dimensional monitoring module transmits the real-time attenuation rate of the storage capacitor calculated by the voltage sampling sub-module to the attenuation rate difference calculation unit of the trend prediction module, transmits the average regional temperature output by the temperature sensing sub-module to the temperature range matching unit of the trend prediction module, and transmits the total pixel working duration recorded by the usage duration recording sub-module to the usage duration range matching unit of the trend prediction module.

[0008] Furthermore, in the intelligent control system for micro-LED display according to the present invention, the state prediction unit includes a trend prediction module; the trend prediction module is provided with a historical database, and the historical database is indexed in three dimensions of "temperature range (such as 20 - 30 °C) - usage duration range (such as 500 - 1000 hours) - attenuation rate difference (the difference between the real-time attenuation rate of the storage capacitor and the real-time attenuation rate of the auxiliary reference capacitor)", and stores typical attenuation trends (including rapid attenuation, stable attenuation, and slow attenuation) under different indexes; the attenuation rate difference calculation unit of the trend prediction module receives the real-time attenuation rate of the storage capacitor transmitted by the multi-dimensional monitoring module and the real-time attenuation rate output by the auxiliary reference capacitor of the drive control unit, and calculates the difference between the two (ΔR = real-time attenuation rate of the storage capacitor - real-time attenuation rate of the auxiliary reference capacitor); the temperature range matching unit receives the average regional temperature transmitted by the multi-dimensional monitoring module and matches the corresponding temperature range in the historical database; the usage duration range matching unit receives the total pixel working duration transmitted by the multi-dimensional monitoring module and matches the corresponding usage duration range in the historical database; the trend prediction module uses a pattern matching algorithm to input the matched temperature range, usage duration range, and attenuation rate difference ΔR into the historical database for matching degree calculation (matching degree = 1 - |real-time value - historical value| / historical value range), selects the typical attenuation trend corresponding to the index with the highest matching degree, and transmits it to the feed-forward compensation module of the dynamic compensation unit as the attenuation trend prediction result of the storage capacitor in the next frame period.

[0009] Furthermore, in the intelligent control system for micro-LED display according to the present invention, the dynamic compensation unit includes a feed-forward compensation module; the feed-forward compensation module receives the attenuation trend prediction result (rapid / stable / slow) of the next frame period output by the trend prediction module, and adjusts the initial voltage amplitude of the column data sub-circuit based on the updated "attenuation trend - voltage increment" mapping relationship (such as increasing the redundant amount by 1.2V during rapid attenuation, and this relationship is optimized through historical calibration data) of the model calibration module of the self-learning calibration unit; simultaneously, based on the updated "attenuation trend - strobe timing offset" mapping relationship (such as advancing the start time of row strobe by 2 μs during rapid attenuation, and this relationship is optimized through historical calibration data) of the model calibration module, adjusts the strobe timing of the row scanning sub-circuit; the feed-forward compensation module transmits the adjusted voltage amplitude parameter to the digital-to-analog converter of the column data sub-circuit (for generating the updated Data signal), and transmits the adjusted strobe timing parameter to the timing controller of the row scanning sub-circuit (for generating the updated Gate signal), and the drive control unit performs the updated voltage writing and row strobe operations.

[0010] Further, in the intelligent control system for micro-LED display according to the present invention, the dynamic compensation unit includes a feedback compensation module; during the pixel emission period (i.e., the TFT conduction stage, corresponding to the period when the row selection signal is at a high level), the feedback compensation module uses the voltage sampling sub-module to monitor the voltage value across the storage capacitor in real time at a frequency of 10 kHz. Based on the difference between the initial write voltage adjusted by the feedforward compensation module (i.e., the voltage amplitude updated by the column data sub-circuit) and the real-time monitored voltage, as well as the monitoring time interval, the real-time attenuation rate of the storage capacitor is calculated; the feedback compensation module obtains the typical threshold corresponding to the attenuation trend in the historical database of the trend prediction module (for example, the rapid attenuation threshold is >10 mV / s, and this threshold is optimized by the model calibration module of the self-learning calibration unit). If the real-time attenuation rate exceeds this threshold, a narrow pulse compensation voltage (pulse width ≤ 1 μs, and the pulse amplitude is determined by the "attenuation trend and compensation pulse amplitude" parameter updated by the model calibration module) is injected into the storage capacitor through the column data sub-circuit during the line blanking period between adjacent row selection signals (the low-level idle period defined by the timing controller of the line scanning sub-circuit); the feedback compensation module transmits the real-time calculated attenuation rate of the storage capacitor to the trend prediction module of the state prediction unit, which is used to compare with the predicted attenuation trend in the historical database to verify the prediction accuracy of the pattern matching algorithm and trigger the parameter update of the model calibration module.

[0011] Further, in the intelligent control system for micro-LED display according to the present invention, the self-learning calibration unit includes a hardware calibration module; the hardware calibration module calls 2 rows × 2 columns of redundant pixels reserved at the edge of the display matrix every 100 hours (sharing the TFT driving circuit process and the storage capacitor preparation process with normal pixels and not participating in the real-time frame display). The reference voltage corresponding to the same standard gray level as the normal pixels (for example, 100% gray level corresponds to 5V, and this voltage is the same as the initial voltage signal written by the column data sub-circuit to the normal pixels) is written into the storage capacitors of the redundant pixels through the column data sub-circuit of the drive control unit; the hardware calibration module collects the actual emission brightness of the redundant pixels during the emission period of the redundant pixels (i.e., the TFT conduction stage when the line scanning sub-circuit selects the redundant pixels) through an array of optical sensors integrated above the color filter layer and corresponding to each pixel unit (one sensor is set directly above each redundant pixel); the hardware calibration module calculates the deviation value between the actual brightness and the preset standard brightness (for example, 400 cd / m², which is the target brightness of normal pixels in the standard environment), and classifies and stores it according to the interval of the real-time attenuation rate of the storage capacitor calculated by the voltage sampling sub-module (such as 5 - 10 mV / s, 10 - 15 mV / s). After establishing the "attenuation rate and brightness deviation" mapping table, it is transmitted to the feedforward compensation module of the dynamic compensation unit (for correcting the "attenuation trend and voltage increment" mapping relationship) and the feedback compensation module (for correcting the "attenuation trend and compensation pulse amplitude" parameter) through the system bus.

[0012] Furthermore, in the intelligent control system for micro-LED display according to the present invention, the self-learning calibration unit includes a model calibration module. The model calibration module obtains the "attenuation rate - brightness deviation" mapping table established by the hardware calibration module. For each attenuation rate interval in the table (such as 5 - 10 mV / s), it extracts the corresponding temperature value (the average regional temperature from the temperature sensing sub-module of the state prediction unit), the usage duration (the total pixel working duration from the usage duration recording sub-module), the attenuation rate (the real-time attenuation rate from the voltage sampling sub-module), and the brightness deviation value (the calculation result from the hardware calibration module), and compares the brightness deviation with a preset threshold (such as ±20 cd / m², which is the maximum brightness fluctuation allowed for normal display). If the brightness deviation exceeds the threshold, it corrects the "attenuation trend - voltage increment" mapping relationship of the feed-forward compensation module of the dynamic compensation unit (such as originally increasing by 1.2 V for rapid attenuation → adjusted to 1.5 V) or the parameter of the "attenuation trend - compensation pulse amplitude" of the feedback compensation module (such as originally 0.3 V → 0.4 V); simultaneously, it adds the extracted temperature, usage duration, attenuation rate, and brightness deviation data to the historical database of the state prediction unit, updates the boundary values of the three-dimensional index of "temperature interval - usage duration interval - attenuation rate difference" based on the distribution characteristics of the new data (such as the minimum / maximum temperature value, the concentrated interval of the attenuation rate) (such as originally 20 - 30 °C → adjusted to 18 - 32 °C), and re-statistics the typical attenuation trends under each index (such as originally the rapid attenuation threshold > 10 mV / s → adjusted to > 8 mV / s, based on the attenuation rate distribution of more than 80% of the samples in the new data), so as to optimize the prediction accuracy of the pattern matching algorithm of the trend prediction module.

[0013] Furthermore, in the intelligent control system for micro-LED display according to the present invention, the auxiliary reference capacitor is integrated in the edge calibration area of the non-display area of the display matrix (physically isolated from the pixel units in the display area and not connected to the LED diodes), and is prepared with the same batch of dielectric materials (such as silicon nitride thin films of the same thickness) and the same process parameters (such as the same magnetron sputtering deposition temperature, the same lithography accuracy of the electrode patterning) as the storage capacitor. A micro-current monitoring circuit (the same as the sampling circuit of the voltage sampling sub-module) is connected in parallel at both ends of the auxiliary reference capacitor to collect its leakage current in real time and calculate the attenuation rate (attenuation rate = leakage current × capacitor plate area / dielectric constant). This attenuation rate data is transmitted to the attenuation rate difference calculation unit of the trend prediction module of the state prediction unit through an independent data bus. Since there is no LED diode load in the calibration area (only including a capacitor structure), the attenuation rate data of the auxiliary reference capacitor can be used as a reference value without display load interference to calculate the difference between the real-time attenuation rate of the storage capacitor and the real-time attenuation rate of the auxiliary reference capacitor (ΔR = real-time attenuation rate of the storage capacitor - real-time attenuation rate of the auxiliary reference capacitor).

[0014] Further, in the micro-LED display intelligent control system of the present invention, the temperature sensing sub-module is a thin-film thermocouple integrated in the pixel gap area of the TFT substrate (one is set at the center position of each 2×2 pixel unit, and a total of m×n thermocouples are set, where m and n are the number of rows and columns of the display matrix). Its hot end is in contact with the TFT substrate of the pixel unit (for sensing the Joule heat when the pixel works), and the cold end is connected to the constant temperature reference point at the edge of the substrate (for eliminating the influence of ambient temperature fluctuations); the thin-film thermocouple collects the temperature values of each distribution point in real time (such as the temperature T1 at point 1 and the temperature T2 at point 2). After converting the analog temperature signal into a digital signal through the analog-to-digital conversion circuit, the multi-dimensional monitoring module of the state prediction unit eliminates outliers (such as extreme values exceeding the temperature tolerance upper limit of the substrate material), and then calculates the average temperature of the remaining valid data (T_avg=(T1+T2+…+Tk) / k, k≤m×n), which is output as the regional average temperature of the display matrix to the temperature range matching unit of the trend prediction module; the temperature range matching unit compares T_avg with the boundary values of each temperature range in the historical database (such as 20-30°C, 30-40°C) to determine the real-time temperature range (such as 25°C belongs to the 20-30°C range).

[0015] Advantages of the present invention; Through the hierarchical cooperation of the drive control unit, the state prediction unit, the dynamic compensation unit and the self-learning calibration unit, the present invention constructs an adaptive control link of "monitoring - prediction - compensation - calibration", effectively solving the coupling failure problem of the storage capacitor voltage maintenance stability and the dynamic brightness regulation consistency: the drive control unit provides the row selection duration, the initial voltage amplitude and the auxiliary reference capacitor reference attenuation rate through the row scanning sub-circuit, the column data sub-circuit and the double capacitor module, providing the original data support for subsequent monitoring and prediction; the state prediction unit collects the residual voltage of the storage capacitor, the regional temperature and the total pixel working duration in real time through the multi-dimensional monitoring module (including the voltage sampling sub-module, the temperature sensing sub-module and the usage duration recording sub-module), calculates the attenuation rate difference in combination with the reference data of the drive control unit, and predicts the attenuation trend by using the three-dimensional index and the pattern matching algorithm of the historical database, providing a forward-looking basis for dynamic compensation; the dynamic compensation unit adjusts the initial voltage amplitude and the gating timing in advance through the feed-forward compensation module, and combines the feedback compensation module to monitor and inject the compensation voltage in real time during the pixel emission period to offset the uncertainty of the storage capacitor attenuation; the self-learning calibration unit establishes a "attenuation rate - brightness deviation" mapping table through the hardware calibration module, and the model calibration module corrects the dynamic compensation parameters based on this table and updates the historical database of the state prediction unit to optimize the prediction accuracy. Through the above technical means, the system can actively cope with the performance drift of the storage capacitor caused by environmental temperature changes and device aging, maintain the matching of the TFT conduction time and the voltage maintenance period, and finally achieve the brightness uniformity of the same gray level area of the display screen. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on the drawings without creative efforts.

[0017] Figure 1 It is a system architecture diagram of the micro-LED display intelligent control system provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention will be described in detail below with reference to the drawings. In order to better understand the objectives of the present invention, the present invention will be further described in detail below.

[0019] Please refer to Figure 1, the micro-LED display intelligent control system provided by the present invention includes a drive control unit, a state prediction unit, a dynamic compensation unit, and a self-learning calibration unit; the drive control unit outputs the row selection duration (high-level duration) through the row scanning sub-circuit, outputs the initial voltage amplitude of the storage capacitor through the column data sub-circuit, and outputs the real-time attenuation rate to the state prediction unit through the auxiliary reference capacitor; the state prediction unit includes a multi-dimensional monitoring module (composed of a voltage sampling sub-module, a temperature sensing sub-module, and a usage duration recording sub-module), which collects the residual voltage of the storage capacitor through the voltage sampling sub-module and calculates the real-time attenuation rate, collects the average regional temperature through the temperature sensing sub-module, accumulates the total working duration of the pixels through the usage duration recording sub-module, combines the initial voltage, row selection duration, and auxiliary reference capacitor attenuation rate output by the drive control unit, and uses the historical database and pattern matching algorithm in the trend prediction module to predict the attenuation trend of the storage capacitor, and transmits the prediction result to the dynamic compensation unit; the dynamic compensation unit adjusts the voltage amplitude of the column data sub-circuit and the gating timing of the row scanning sub-circuit through the feed-forward compensation module, and monitors the real-time attenuation rate of the storage capacitor during pixel emission through the feedback compensation module and feeds it back to the trend prediction module of the state prediction unit for model verification; the self-learning calibration unit generates a "decay rate and brightness deviation" mapping table by calling the redundant pixels at the edge of the display matrix through the hardware calibration module, and the model calibration module updates the prediction parameters of the historical database of the state prediction unit based on the decay rate and brightness deviation mapping table, and synchronously corrects the compensation parameters of the feed-forward module and feedback module of the dynamic compensation unit to form a closed-loop control link of "driving data providing - state prediction - dynamic compensation - calibration optimization", so as to solve the problem of uneven brightness in the same gray-scale area of the display screen caused by the coupling failure between the voltage maintenance stability of the storage capacitor and the dynamic brightness regulation consistency.

[0020] The micro-LED display intelligent control system solves the problem of uneven brightness in the same gray-scale area of the display screen caused by the coupling failure between the voltage maintenance stability of the storage capacitor and the dynamic brightness regulation consistency through the collaborative work of the drive control unit, the state prediction unit, the dynamic compensation unit, and the self-learning calibration unit.

[0021] The drive control unit, as the basic signal transmission module of the system, includes a row scanning sub-circuit, a column data sub-circuit, and a dual-capacitor module. The row scanning sub-circuit generates a row strobe signal (Gate signal) by a timing controller, and outputs a high-level signal row by row in the order of row addresses (such as the 1st row → the 2nd row →... → the nth row), activating the TFT gates of the corresponding row pixel units to make the TFTs enter the conduction state; meanwhile, the sub-circuit outputs the high-level duration of the row strobe signal (row strobe duration) to the state prediction unit for subsequent calculation of the voltage decay rate. The column data sub-circuit generates a column data signal (Data signal) by a data driver, and only during the period when the corresponding row strobe signal is at a high level (i.e., the TFT conduction stage), writes an initial voltage signal corresponding to the image gray level (such as 5V for 100% gray level and 2.5V for 50% gray level) to the storage capacitors of all pixel units in that row; the sub-circuit synchronously outputs the initial voltage amplitude to the state prediction unit as a reference value for voltage decay calculation. The dual-capacitor module includes a storage capacitor connected in parallel between the TFT gate and the ground (for temporarily storing voltage signals to maintain TFT conduction) and an auxiliary reference capacitor integrated in the non-display area of the display matrix (prepared with the same dielectric material and electrode process as the storage capacitor, without LED diode load interference); the auxiliary reference capacitor collects its own leakage current in real time through a micro-current monitoring circuit and calculates the decay rate, which is transmitted to the state prediction unit as a reference value for comparing the actual decay characteristics of the storage capacitor.

[0022] The state prediction unit provides a forward-looking basis for dynamic compensation through multi-dimensional monitoring and trend prediction. The multi-dimensional monitoring module includes three sub-modules: The voltage sampling sub-module collects the residual voltage value through a micro-current sampling circuit connected in parallel across the storage capacitor at the end of each row strobe period (the moment when the row strobe signal jumps from high level to low level), and calculates the attenuation rate of the storage capacitor for the current period based on the initial voltage amplitude and the row strobe duration provided by the drive control unit (attenuation rate = (initial voltage amplitude - residual voltage value) / row strobe duration); The temperature sensing sub-module integrates thin-film thermocouples in the pixel gap area of the TFT substrate (the center position of each 2×2 pixel unit), and collects the temperature values of each distribution point in real time (such as the temperature T1 at point 1 and the temperature T2 at point 2). After analog-to-digital conversion, the average temperature of the valid data is calculated (averaged after removing outliers) and used as the regional temperature parameter of the display matrix; The usage duration recording sub-module configures a non-volatile counter (such as an OTP memory) for each pixel unit, and only accumulates and records the conduction time during the period when the TFT is conducting (the period when the row strobe signal is at high level). When the system starts, the total working duration of the pixel unit is read and updated. The trend prediction module receives the attenuation rate of the storage capacitor, the average regional temperature, and the total pixel working duration output by the multi-dimensional monitoring module, and combines the attenuation rate of the auxiliary reference capacitor provided by the drive control unit to calculate the difference in attenuation rates between the two (ΔR = attenuation rate of the storage capacitor - attenuation rate of the auxiliary reference capacitor); Through the pattern matching algorithm of the historical database (storing typical attenuation trends in a three-dimensional index of "temperature range - usage duration range - attenuation rate difference", such as fast / stable / slow), the trend corresponding to the index with the highest matching degree is selected as the prediction result and transmitted to the dynamic compensation unit.

[0023] The dynamic compensation unit cancels the uncertainty of the storage capacitor attenuation in two stages of feedforward and feedback based on the predicted trend. The feedforward compensation module receives the trend prediction result (fast / stable / slow), and adjusts the initial voltage amplitude of the column data sub-circuit by using the "attenuation trend - voltage increment" mapping relationship optimized by the self-learning calibration unit (such as increasing the redundant amount by 1.2V during fast attenuation), and at the same time adjusts the gating timing of the row scanning sub-circuit by using the "attenuation trend - gating timing offset" mapping relationship (such as advancing the starting time of row gating by 2μs during fast attenuation); the adjusted parameters are respectively transmitted to the digital-to-analog converter of the column data sub-circuit (to generate an updated Data signal) and the timing controller of the row scanning sub-circuit (to generate an updated Gate signal), and the updated voltage writing and row gating operations are executed by the drive control unit. The feedback compensation module, during the pixel emission period (TFT conduction stage), monitors the voltage value across the storage capacitor in real time at a frequency of 10kHz through the voltage sampling sub-module, and calculates the real-time attenuation rate based on the difference between the initial voltage after feedforward adjustment and the currently monitored voltage, and the monitoring time interval; if this rate exceeds the typical threshold corresponding to the trend in the historical database of the trend prediction module (such as the fast attenuation threshold is >10mV / s), then a narrow pulse compensation voltage (pulse width ≤1μs, amplitude determined by the calibration parameter) is injected into the storage capacitor through the column data sub-circuit during the blanking period of adjacent row gating (idle period when the row gating signal is low), and the real-time attenuation rate is fed back to the trend prediction module to verify the prediction accuracy.

[0024] The self-learning calibration unit enhances the system robustness in a closed-loop manner through hardware verification and model optimization. The hardware calibration module calls the 2 rows × 2 columns redundant pixels reserved at the edge of the display matrix every 100 hours (sharing the TFT driving circuit and storage capacitor fabrication process with normal pixels and not participating in the current frame display). The column data sub-circuit of the drive control unit writes the same standard gray reference voltage as the normal pixels (e.g., 5V corresponding to 100% gray) to the storage capacitors of the redundant pixels. The array-type optical sensors integrated above the color film layer (one is set directly above each redundant pixel) collect the actual brightness during the emission of the redundant pixels, calculate the deviation value from the preset standard brightness (e.g., 400 cd / m²), and classify and store it according to the storage capacitor decay rate intervals (such as 5 - 10 mV / s, 10 - 15 mV / s), establishing a "decay rate - brightness deviation" mapping table. Based on this mapping table, the model calibration module extracts the corresponding temperature, usage duration, decay rate, and brightness deviation data. If the brightness deviation in a certain interval exceeds the threshold (e.g., ±20 cd / m²), it corrects the "decay trend - voltage increment" mapping relationship of the feedforward module or the "decay trend - compensation pulse amplitude" parameter of the feedback module of the dynamic compensation unit; simultaneously adds the new data to the historical database of the state prediction unit, updates the boundary values of the three-dimensional index of "temperature interval - usage duration interval - decay rate difference" (such as the original 20 - 30 °C → 18 - 32 °C) and the typical thresholds of the corresponding trends (such as the original rapid decay threshold > 10 mV / s → > 8 mV / s), and optimizes the prediction accuracy of the pattern matching algorithm.

[0025] Through the hierarchical data interaction and strategy iteration among the above modules, the system forms an adaptive control link of "driving data provision - state prediction - dynamic compensation - calibration optimization", actively coping with the performance drift of storage capacitors caused by environmental temperature changes, device aging, etc., and solving the problems of TFT conduction time deviation and uneven brightness in the gray scale area caused by unstable voltage maintenance.

[0026] Specifically, for the micro-LED display intelligent control system of the present invention, the drive control unit includes a row scanning sub-circuit, a column data sub-circuit, and a dual-capacitor module; the row scanning sub-circuit generates a row strobe signal (Gate signal) by a timing controller, and outputs a high-level signal row by row in the order of row addresses (such as the 1st row → the 2nd row → … → the nth row), activating the TFT gates of the corresponding row pixel units to enter the conduction state; the column data sub-circuit generates a column data signal (Data signal) by a data driver, and only during the period when the corresponding row strobe signal is high level (i.e., the TFT conduction stage), writes an initial voltage signal corresponding to the image gray level (such as 5V for 100% gray level and 2.5V for 50% gray level) to the storage capacitors of all pixel units in this row; the dual-capacitor module includes a storage capacitor connected in parallel between the TFT gate and the ground (for temporarily storing the voltage signal to maintain the TFT conduction) and an auxiliary reference capacitor integrated in the non-display area of the display matrix (prepared with the same dielectric material and electrode process as the storage capacitor, for providing reference attenuation rate data without display load interference); the row scanning sub-circuit outputs the row strobe duration (high-level duration) to the voltage sampling sub-module of the state prediction unit, the column data sub-circuit outputs the initial voltage amplitude of the storage capacitor to the voltage sampling sub-module of the state prediction unit, and the auxiliary reference capacitor outputs the real-time attenuation rate to the trend prediction module of the state prediction unit through a micro-current monitoring circuit.

[0027] The drive control unit is the basic signal transmission module of the micro-LED display intelligent control system, and its core function is to provide the original data support for subsequent state prediction and dynamic compensation, which is specifically realized by the cooperation of the row scanning sub-circuit, the column data sub-circuit, and the dual-capacitor module.

[0028] The core component of the row scanning sub-circuit is the timing controller, which generates a periodic row strobe signal (Gate signal) through an internal clock signal. This signal outputs high-level pulses row by row in the order of row addresses (such as the 1st row → the 2nd row → … → the nth row): when the Gate signal of the ith row is at a high level, the TFT gates of all pixel units in this row are activated, and the TFT enters the conduction state because the gate voltage reaches the conduction threshold. The row scanning sub-circuit synchronously records the duration of the high-level signal (i.e., the row strobe duration), and transmits this duration as a key timing parameter to the voltage sampling sub-module of the state prediction unit for calculating the voltage attenuation rate of the storage capacitor in the future. For example, if the high-level duration of the Gate signal of the 3rd row is 10 μs, this duration will be used by the voltage sampling sub-module to calculate the voltage attenuation amount of the storage capacitor in this row within 10 μs.

[0029] The column data sub - circuit is driven by a data driver. Its function is to convert the image grayscale signal into a corresponding voltage signal and write it into the storage capacitor. The data driver receives grayscale data (such as 8 - bit grayscale values 0 - 255) from the display control chip, and generates an analog voltage signal linearly corresponding to the grayscale value through an internal digital - to - analog conversion circuit (for example, 100% grayscale corresponds to 5V, 50% grayscale corresponds to 2.5V). This voltage signal is transmitted to the storage capacitors of all pixel units in that row through the column data line only during the period when the corresponding row Gate signal is high (i.e., the TFT conduction stage of that row). The column data sub - circuit synchronously outputs the initial voltage amplitude (such as 5V or 2.5V) written to the voltage sampling sub - module of the state prediction unit, serving as the reference value for calculating the voltage decay of the storage capacitor, so as to achieve a clear initial reference for calculating the subsequent decay rate.

[0030] The dual - capacitor module includes a storage capacitor and an auxiliary reference capacitor, which jointly provide voltage decay characteristic data for state prediction. The storage capacitor is connected in parallel between the TFT gate and the ground. Its function is to temporarily store the voltage signal written by the column data sub - circuit, and keep the TFT gate voltage to maintain its conduction state, thereby controlling the current magnitude and light - emitting brightness of the LED diode. The auxiliary reference capacitor is integrated in the non - display area of the display matrix (such as the calibration area at the edge of the matrix), and uses the same batch of dielectric materials (such as silicon nitride thin film) and exactly the same preparation process as the storage capacitor (such as the same magnetron sputtering deposition temperature, electrode patterning lithography accuracy), so as to achieve highly consistent intrinsic decay characteristics of the two. Since the auxiliary reference capacitor is not connected to the LED diode (no display load), its decay is only caused by the leakage current of the dielectric material itself. The leakage current can be collected in real - time through a parallel micro - current monitoring circuit (the same model as the sampling circuit of the voltage sampling sub - module), and the decay rate (such as 2mV / s) can be calculated based on parameters such as the capacitor plate area and dielectric constant. This decay rate is used as a reference value without load interference and is transmitted to the trend prediction module of the state prediction unit through an independent data bus for comparing the actual decay characteristics of the storage capacitor (the decay rate of the storage capacitor includes the influence of the load due to its connection to the LED diode).

[0031] The three sub - modules of the drive control unit form a basic signal link through clear timing coordination and data interaction: the row scanning sub - circuit activates the TFT row by row through the Gate signal, the column data sub - circuit writes voltage during the TFT conduction period, and the dual - capacitor module provides the actual storage voltage and reference decay data respectively; the row selection duration, the initial voltage amplitude, and the decay rate of the auxiliary reference capacitor are transmitted to the corresponding modules of the state prediction unit as key parameters, providing necessary inputs for subsequent multi - dimensional monitoring and trend prediction. This design realizes the accurate perception of the voltage decay characteristics of the storage capacitor by the system, which is the premise for realizing dynamic compensation and self - learning calibration.

[0032] Specifically, in the micro-LED display intelligent control system of the present invention, the state prediction unit includes a multi-dimensional monitoring module; the multi-dimensional monitoring module includes a voltage sampling sub-module, a temperature sensing sub-module, and a usage duration recording sub-module; at the end of each row strobe cycle (i.e., the moment when the row strobe signal jumps from high level to low level), the voltage sampling sub-module collects the residual voltage value of the storage capacitor through a micro-current sampling circuit connected in parallel across the two ends of the storage capacitor, and calculates the attenuation rate of the storage capacitor within the real-time period based on the initial voltage amplitude output by the column data sub-circuit of the drive control unit and the row strobe duration output by the row scanning sub-circuit (attenuation rate = (initial voltage amplitude - residual voltage value) / row strobe duration); the temperature sensing sub-module integrates thin-film thermocouples in the pixel gap area of the TFT substrate (at the center position of each 2×2 pixel unit), and collects the temperature values of each distribution point in real time (such as the temperature T1 at point 1 and the temperature T2 at point 2). After converting the analog temperature signal into a digital signal through an analog-to-digital conversion circuit, the average temperature of all distribution points (T_avg = (T1 + T2 + … + Tn) / n) is calculated as the average regional temperature; the usage duration recording sub-module configures a non-volatile counter (such as an OTP memory) for each pixel unit, and only accumulatively records the conduction time during the period when the TFT is conducting (i.e., the period when the row strobe signal is at high level). When the system starts, the total working duration of the pixel unit is read and updated; the multi-dimensional monitoring module transmits the real-time attenuation rate of the storage capacitor calculated by the voltage sampling sub-module to the attenuation rate difference calculation unit of the trend prediction module, transmits the average regional temperature output by the temperature sensing sub-module to the temperature range matching unit of the trend prediction module, and transmits the total working duration of the pixel recorded by the usage duration recording sub-module to the usage duration range matching unit of the trend prediction module.

[0033] The state prediction unit is the core perception and prediction module of the micro-LED display intelligent control system. It collects the operation state data of the storage capacitor in real time through the multi-dimensional monitoring module, and predicts the attenuation trend in combination with historical data, providing a decision-making basis for subsequent dynamic compensation. As the basic component of the state prediction unit, the multi-dimensional monitoring module works in cooperation with the voltage sampling sub-module, the temperature sensing sub-module, and the usage duration recording sub-module, and obtains key parameters from three dimensions: voltage attenuation, ambient temperature, and device aging.

[0034] The voltage sampling submodule quantifies the voltage-maintaining capability of the storage capacitor. This module activates at the end of each row-strobe cycle (i.e., the instant the row-strobe signal output by the row-scan subcircuit transitions from high to low). It collects the residual voltage value via a micro-current sampling circuit (the same model as the monitoring circuit for the auxiliary reference capacitor in the driver control unit) connected in parallel across the storage capacitor. The micro-current sampling circuit utilizes a high input impedance design, ensuring that the sampling process has negligible impact on the storage capacitor voltage. The collected residual voltage value, along with the initial voltage amplitude output by the column data subcircuit of the driver control unit (e.g., 5V for 100% grayscale) and the row-strobe duration output by the row-scan subcircuit (e.g., 10μs), serves as input for the calculation. The voltage decay rate of the row storage capacitor during the current cycle is calculated using the formula "decay rate = (initial voltage amplitude - residual voltage) / row-strobe duration" to determine the voltage decay rate (e.g., 3mV / μs) for the row storage capacitor during the current cycle. This rate directly reflects the storage capacitor's ability to maintain a voltage signal. A higher rate indicates faster voltage decay due to leakage current or dielectric drift.

[0035] The temperature sensing submodule is used to obtain the regional temperature distribution of the display matrix. This module integrates thin-film thermocouples in the pixel gap area of the TFT substrate (at the center of each 2×2 pixel unit). The hot end of each thermocouple is in direct contact with the TFT substrate (to sense Joule heating during pixel operation), and the cold end is connected to a constant-temperature reference point at the edge of the substrate (to eliminate interference from ambient temperature fluctuations). The thin-film thermocouples output a real-time analog voltage signal proportional to temperature (e.g., 10μV / °C). This signal is converted into a digital temperature value (e.g., point 1 temperature T1 = 25°C, point 2 temperature T2 = 28°C) via an analog-to-digital converter (ADC). To improve data reliability, the multi-dimensional monitoring module pre-processes the collected temperature values. After removing outliers that exceed the upper temperature limit of the substrate material (e.g., 120°C), the average value of the remaining valid data (T_avg = (T1+T2+…+Tn) / n) is calculated as the regional average temperature of the display matrix (e.g., 26.5°C). This parameter is used to assess the impact of temperature on the dielectric properties of storage capacitors (increased temperature generally increases leakage current and accelerates voltage decay).

[0036] The usage duration recording submodule tracks the cumulative operating time of each pixel unit. Each pixel unit is equipped with a non-volatile counter (such as an OTP memory). This counter accumulates time only during the TFT on-time (i.e., when the row select signal is high). When the row select signal is high, the counter increments based on the system clock (e.g., 1 MHz); when the row select signal transitions to a low level, the counter pauses. At system startup, the memory controller reads the counter values of all pixel units and updates them to the current total operating time (e.g., the cumulative on-time of a pixel is 1000 hours). This parameter is used to quantify device aging (performance drift of the TFT and storage capacitor is generally positively correlated with cumulative operating time).

[0037] The multi-dimensional monitoring module transmits the output data of the three sub-modules to the corresponding units of the trend prediction module respectively: the real-time attenuation rate of the storage capacitor calculated by the voltage sampling sub-module is transmitted to the attenuation rate difference calculation unit, which is compared with the reference attenuation rate of the auxiliary reference capacitor of the drive control unit to obtain the attenuation rate difference reflecting the influence of the load; the average regional temperature output by the temperature sensing sub-module is transmitted to the temperature range matching unit, which is matched with the temperature ranges in the historical database (such as 20 - 30 °C, 30 - 40 °C) to determine the current temperature environment; the total pixel working hours recorded by the usage duration recording sub-module are transmitted to the usage duration range matching unit, which is matched with the usage duration ranges in the historical database (such as 500 - 1000 hours, 1000 - 1500 hours) to determine the device aging stage. The parameters of these three dimensions jointly constitute the input of trend prediction, providing multi-source data support for the subsequent pattern matching algorithm and enabling the prediction of the attenuation trend of the storage capacitor to be more in line with the actual operating state.

[0038] Specifically, in the intelligent regulation system for micro-LED display of the present invention, the state prediction unit includes a trend prediction module; the trend prediction module is provided with a historical database, and the historical database is indexed in three dimensions of "temperature range (such as 20 - 30 °C) - usage duration range (such as 500 - 1000 hours) - attenuation rate difference (the difference between the real-time attenuation rate of the storage capacitor and the real-time attenuation rate of the auxiliary reference capacitor)", and stores typical attenuation trends (including rapid attenuation, stable attenuation, slow attenuation) under different indexes; the attenuation rate difference calculation unit of the trend prediction module receives the real-time attenuation rate of the storage capacitor transmitted by the multi-dimensional monitoring module and the real-time attenuation rate output by the auxiliary reference capacitor of the drive control unit, and calculates the difference between the two (ΔR = real-time attenuation rate of the storage capacitor - real-time attenuation rate of the auxiliary reference capacitor); the temperature range matching unit receives the average regional temperature transmitted by the multi-dimensional monitoring module and matches the corresponding temperature range in the historical database; the usage duration range matching unit receives the total pixel working hours transmitted by the multi-dimensional monitoring module and matches the corresponding usage duration range in the historical database; the trend prediction module uses a pattern matching algorithm to input the matched temperature range, usage duration range and attenuation rate difference ΔR into the historical database for matching degree calculation (matching degree = 1 - |real-time value - historical value| / historical value range), and selects the typical attenuation trend corresponding to the index with the highest matching degree as the prediction result of the attenuation trend of the storage capacitor in the next frame period and transmits it to the feed-forward compensation module of the dynamic compensation unit.

[0039] The trend prediction module is the core decision-making component of the state prediction unit. It predicts the attenuation trend of the storage capacitor by matching and analyzing historical data with real-time monitoring parameters, providing a control basis for the dynamic compensation unit. The function of this module relies on the three-dimensional index structure of the historical database, the matching calculation of multi-source parameters, and the application of pattern matching algorithms. The specific process is as follows: The historical database forms the knowledge foundation for trend forecasting and is constructed using a three-dimensional index: "temperature range - usage duration range - decay rate difference." Temperature ranges are based on the common operating temperature ranges of the display matrix (e.g., 20-30°C, 30-40°C, 40-50°C), each corresponding to the typical temperature characteristics of the storage capacitor's dielectric material. Usage duration ranges are based on typical stages of device aging (e.g., 500-1000 hours, 1000-1500 hours, 1500-2000 hours), each corresponding to the degree of aging of the TFT and storage capacitor. The decay rate difference (ΔR = real-time decay rate of the storage capacitor - real-time decay rate of the auxiliary reference capacitor) reflects the additional decay of the storage capacitor due to the connected LED diode load (e.g., ΔR > 5mV / μs indicates significantly accelerated decay). Each three-dimensional index combination in the database (such as "20-30℃-500-1000 hours-ΔR=3-6mV / μs") stores the typical decay trend (fast / stable / slow) under that condition. These trends are derived from historical data from previous system operations (for example, if 80% of samples show rapid decay under that index, it is marked as a fast trend).

[0040] The decay rate difference calculation unit is responsible for quantifying the load effect on the storage capacitor. This unit receives the real-time decay rate of the storage capacitor (e.g., 4mV / μs) transmitted by the multi-dimensional monitoring module and the real-time decay rate output by the auxiliary reference capacitor of the drive control unit (e.g., 1mV / μs), and calculates the difference between the two (ΔR = 4mV / μs - 1mV / μs = 3mV / μs). This difference directly reflects the additional attenuation of the storage capacitor caused by the connected LED diode (load) and is a key parameter for distinguishing intrinsic attenuation (auxiliary reference capacitor) from load effects (storage capacitor).

[0041] The temperature range matching unit compares the average regional temperature output by the multi-dimensional monitoring module (e.g., 25°C) with the temperature range boundaries in the historical database (e.g., 20-30°C, 30-40°C) to determine the current temperature range (e.g., 25°C belongs to the 20-30°C range). This matching process is achieved through numerical comparison, ensuring that subsequent analysis is based on the historical data that most closely matches the current temperature environment.

[0042] The usage duration interval matching unit compares the total pixel working duration recorded by the multi-dimensional monitoring module (such as 800 hours) with the usage duration intervals in the historical database (such as 500 - 1000 hours, 1000 - 1500 hours) to determine the current aging stage (800 hours belongs to the 500 - 1000 hours interval). This matching process is also completed through numerical comparison, enabling the analysis to be based on historical data that is closest to the current device aging degree.

[0043] The pattern matching algorithm is the core logic of trend prediction. After obtaining the temperature interval (20 - 30 °C), usage duration interval (500 - 1000 hours), and attenuation rate difference (ΔR = 3 mV / μs), the algorithm traverses all three-dimensional indexes in the historical database that match these three parameters (such as "20 - 30 °C - 500 - 1000 hours - ΔR = 3 - 6 mV / μs"), and calculates the matching degree between the current parameters and each index (matching degree = 1 - |current value - historical value| / historical value range). For example, if the historical ΔR range of a certain index is 3 - 6 mV / μs and the current ΔR = 3 mV / μs, then the matching degree = 1 - |3 - 4.5| / 3 = 0.5 (assuming the historical value takes the middle value of 4.5 mV / μs). The algorithm selects the index with the highest matching degree (such as the index with a matching degree of 0.8), and uses its corresponding typical attenuation trend (such as rapid attenuation) as the prediction result of the attenuation trend of the storage capacitor in the next frame period.

[0044] Finally, the prediction result (such as "rapid attenuation") is transmitted to the feedforward compensation module of the dynamic compensation unit through the system bus, providing a forward-looking basis for adjusting the initial voltage amplitude and row strobe timing. Through the precise matching of multi-dimensional parameters and the statistical analysis of historical data, this process realizes the quantitative prediction of the attenuation trend of the storage capacitor, enabling the dynamic compensation strategy to actively address potential voltage retention instability problems.

[0045] Specifically, in the micro-LED display intelligent control system of the present invention, the dynamic compensation unit includes a feed-forward compensation module; the feed-forward compensation module receives the next frame cycle attenuation trend prediction result (fast / stable / slow) output by the trend prediction module, and based on the "attenuation trend and voltage increment" mapping relationship updated by the model calibration module of the self-learning calibration unit (such as increasing the redundant amount by 1.2V during fast attenuation, and this relationship is optimized through historical calibration data), adjusts the initial voltage amplitude of the column data sub-circuit; simultaneously, based on the "attenuation trend and strobe timing offset" mapping relationship updated by the model calibration module (such as advancing the row strobe start time by 2μs during fast attenuation, and this relationship is optimized through historical calibration data), adjusts the strobe timing of the row scanning sub-circuit; the feed-forward compensation module transmits the adjusted voltage amplitude parameter to the digital-to-analog converter of the column data sub-circuit (for generating the updated Data signal), and transmits the adjusted strobe timing parameter to the timing controller of the row scanning sub-circuit (for generating the updated Gate signal), and the drive control unit performs the updated voltage writing and row strobe operations.

[0046] The feed-forward compensation module of the dynamic compensation unit is a key component for the system to achieve active control. Its core function is to adjust the voltage writing and row strobe timing of the drive control unit in advance based on the trend prediction result to offset the problem of insufficient voltage maintenance caused by the attenuation of the storage capacitor. The working process of this module closely depends on the trend prediction result and the mapping relationship optimized by self-learning calibration. The specific implementation steps are as follows: The feed-forward compensation module first receives the next frame cycle attenuation trend prediction result (represented in the form of an enumerated value, such as "fast attenuation", "stable attenuation", "slow attenuation") transmitted by the trend prediction module through the system bus. This result directly reflects the voltage attenuation characteristics of the storage capacitor in the next frame cycle: "fast attenuation" means that the voltage of the storage capacitor will rapidly decrease due to an increase in leakage current or dielectric property drift; "stable attenuation" means that the voltage attenuation rate is consistent with the design expectation; "slow attenuation" means that the voltage attenuation rate is lower than expected (possibly due to a decrease in environmental temperature or a slowdown in device aging).

[0047] Based on the received attenuation trend, the feedforward compensation module calls the internally stored "attenuation trend - voltage increment" mapping relation table. This table is regularly updated by the model calibration module of the self - learning calibration unit (optimized, for example, after every 5 hardware calibrations), and stores the voltage increment values corresponding to different attenuation trends (such as rapid attenuation → +1.2V, stable attenuation → +0.5V, slow attenuation → 0V). For example, if the prediction result is "rapid attenuation", the module looks up the corresponding increment value +1.2V in the table and superimposes it on the original initial voltage amplitude of the column data sub - circuit of the drive control unit (such as 5V corresponding to 100% gray - scale), obtaining the adjusted voltage amplitude (5V + 1.2V = 6.2V). This increment compensates for the attenuation loss of the storage capacitor in subsequent cycles by increasing the voltage redundancy, enabling the TFT gate voltage to be maintained above the effective threshold within the target conduction duration.

[0048] Synchronously, the feedforward compensation module calls another "attenuation trend - strobe timing offset" mapping relation table optimized by the model calibration module (such as rapid attenuation → advance by 2μs, stable attenuation → advance by 1μs, slow attenuation → no offset). If the prediction result is "rapid attenuation", the module obtains the offset - 2μs (the negative sign indicates advance) from the table and adjusts the strobe timing of the row scanning sub - circuit: advancing the rising edge of the row strobe signal (Gate signal) from the original timing point (such as the 10th μs) to the 8th μs, making the start time of TFT conduction earlier than the starting point of the storage capacitor voltage attenuation. This adjustment extends the effective conduction duration of the TFT (the original duration was 10μs, and after adjustment it is 12μs), enabling a more sufficient overlap between the TFT conduction cycle and the effective voltage maintenance cycle of the storage capacitor.

[0049] The adjusted voltage amplitude parameter and strobe timing parameter are transmitted to the corresponding sub - circuits of the drive control unit through different control buses: the voltage amplitude parameter (such as 6.2V) is transmitted to the digital - to - analog converter (DAC) of the column data sub - circuit in the form of a digital signal, and the DAC converts it into an analog voltage signal and outputs it as the new Data signal to the column data line; the strobe timing parameter (such as advance by 2μs) is transmitted to the timing controller of the row scanning sub - circuit in the form of a control instruction, and the controller adjusts the phase of the internal clock signal according to this instruction to generate the updated Gate signal.

[0050] The drive control unit executes the updated operations: the column data sub - circuit writes the adjusted voltage (6.2V) to the storage capacitor through the column data line during the period when the corresponding row strobe signal is high; the row scanning sub - circuit outputs the Gate signal according to the adjusted timing, activating the TFT gate in advance. Through this active adjustment, the storage capacitor obtains higher voltage redundancy in the initial stage, and the TFT conduction time is more matched with the voltage maintenance cycle, effectively offsetting the voltage shortage problem caused by attenuation.

[0051] The design of the feedforward compensation module converts the trend prediction result into specific control parameters through the closed-loop logic of "prediction - adjustment - execution". This enables the system to intervene actively before the voltage decay of the storage capacitor significantly affects the display effect, avoiding the lag problem of traditional methods that rely solely on real-time feedback. It is the core technical means to achieve brightness uniformity in the gray-scale area of the display screen.

[0052] Specifically, in the intelligent control system for micro-LED displays of the present invention, the dynamic compensation unit includes a feedback compensation module. During pixel emission (i.e., the TFT conduction stage, corresponding to the period when the row selection signal is high), the feedback compensation module uses the voltage sampling sub-module to continuously monitor the voltage value across the storage capacitor at a frequency of 10 kHz. Based on the difference between the initial write voltage adjusted by the feedforward compensation module (i.e., the voltage amplitude updated by the column data sub-circuit) and the real-time monitored voltage, as well as the monitoring time interval, it calculates the real-time decay rate of the storage capacitor. The feedback compensation module obtains the typical threshold corresponding to the decay trend in the historical database of the trend prediction module (for example, the rapid decay threshold is >10 mV / s, which is optimized by the model calibration module of the self-learning calibration unit). If the real-time decay rate exceeds this threshold, during the row blanking period between adjacent row selection signals (the low-level idle period defined by the timing controller of the row scanning sub-circuit), the feedback compensation module injects a narrow pulse compensation voltage into the storage capacitor through the column data sub-circuit (pulse width ≤1 μs, and the pulse amplitude is determined by the "decay trend and compensation pulse amplitude" parameter updated by the model calibration module). The feedback compensation module transmits the real-time calculated decay rate of the storage capacitor to the trend prediction module of the state prediction unit, which is used to compare with the predicted decay trend in the historical database to verify the prediction accuracy of the pattern matching algorithm and trigger the parameter update of the model calibration module.

[0053] The feedback compensation module of the dynamic compensation unit is a key supplementary component for the system to achieve precise control. Its core function is to continuously monitor the voltage decay state of the storage capacitor during pixel emission, correct the decay deviation that is not fully offset by the feedforward compensation, and provide data support for verifying the accuracy of the trend prediction model. The working process of this module is closely integrated with real-time monitoring, threshold comparison, and compensation injection. The specific implementation steps are as follows: The monitoring process of the feedback compensation module starts from the pixel lighting stage (i.e., the TFT conduction stage, corresponding to the period when the row selection signal output by the row scanning sub - circuit is at a high level). In this stage, the storage capacitor is controlling the current magnitude and brightness of the LED diode by maintaining the gate voltage of the TFT, and its voltage decay rate directly affects the display effect. The module monitors the voltage value across the storage capacitor in real - time at a frequency of 10 kHz (i.e., once every 100 μs) through the voltage sampling sub - module: the high - input - impedance sampling circuit of the voltage sampling sub - module is connected in parallel across the storage capacitor, enabling the disturbance of the capacitor voltage during the sampling process to be negligible, and the monitoring data is transmitted to the processing unit of the feedback compensation module through a high - speed interface.

[0054] Based on the real - time monitored voltage value, the feedback compensation module calculates the real - time decay rate of the storage capacitor. The calculation inputs include the difference between the initial write voltage adjusted by the feed - forward compensation module (i.e., the voltage amplitude updated by the column data sub - circuit, such as 6.2 V) and the current monitored voltage (such as 5.8 V) (6.2 V - 5.8 V = 0.4 V), and the time interval between two samplings (100 μs). Through the formula "decay rate = voltage difference / time interval", the real - time decay rate in the current cycle is obtained (0.4 V / 100 μs = 4 mV / μs). This rate reflects the actual voltage decay of the storage capacitor during the lighting stage and is a key indicator for evaluating the feed - forward compensation effect.

[0055] The feedback compensation module needs to determine whether the real - time decay rate exceeds the acceptable range. The module extracts the typical threshold (such as 10 mV / μs, which is optimized by the model calibration module of the self - learning calibration unit based on historical calibration data, for example, determined by statistically analyzing the decay rates of 90% of the normal display samples) corresponding to the current decay trend (such as "rapid decay") from the historical database of the trend prediction module. If the real - time decay rate (4 mV / μs) does not exceed the threshold (10 mV / μs), no compensation is triggered; if it exceeds the threshold (such as the real - time rate is 12 mV / μs), the compensation mechanism is activated.

[0056] The injection timing and parameters of the compensation voltage are dynamically determined by the feedback compensation module. The injection timing is selected during the horizontal blanking period between two adjacent row strobe signals (i.e., the idle period when the row strobe signal is low, usually 2 - 5 μs), during which no pixels are in the emitting state, and the injection operation will not interfere with the current frame display. The compensation voltage is in the form of a narrow pulse (pulse width ≤ 1 μs), and its amplitude is determined by the "attenuation trend - compensation pulse amplitude" mapping table updated by the model calibration module (e.g., "rapid attenuation" corresponds to 0.3 V, which is optimized through the "attenuation rate - brightness deviation" data in historical calibration to reduce the brightness deviation to the allowable range after injection). The pulse is generated by the digital - to - analog converter in the column data sub - circuit and transmitted to the storage capacitor through the column data line to supplement the voltage amount lost due to rapid attenuation (e.g., after injecting 0.3 V, the voltage of the storage capacitor rises to 6.1 V).

[0057] After the feedback compensation module completes the compensation, it transmits the attenuation rate of the storage capacitor calculated in real - time (such as 12 mV / μs) to the trend prediction module of the state prediction unit through the system bus. The trend prediction module compares this rate with the previously predicted attenuation trend (such as "rapid attenuation") and calculates the prediction error (e.g., the typical rate corresponding to the predicted trend is 10 mV / μs, the actual rate is 12 mV / μs, and the error is +20%). If the error exceeds the preset threshold (such as ±15%), the model calibration module of the self - learning calibration unit is triggered to update the typical trend threshold and the "attenuation trend - compensation pulse amplitude" mapping table in the historical database (such as adjusting the typical threshold of "rapid attenuation" to 11 mV / μs, or increasing the compensation pulse amplitude to 0.4 V) to improve the accuracy of subsequent prediction and compensation.

[0058] The feedback compensation module uses a closed - loop logic of "real - time monitoring - deviation judgment - dynamic compensation - model verification" to make up for the limitation of feed - forward compensation that only relies on prediction, enabling the system to handle prediction errors caused by environmental mutations or device characteristic discreteness, and further improving the brightness uniformity in the gray - scale area of the display screen. This design combines the forward - looking nature of feed - forward and the real - time nature of feedback, which is the core technical feature for the intelligent control system to achieve high - precision control.

[0059] Specifically, in the micro-LED display intelligent control system of the present invention, the self-learning calibration unit includes a hardware calibration module; the hardware calibration module calls 2 rows × 2 columns of redundant pixels reserved at the edge of the display matrix every 100 hours (sharing the TFT driving circuit process and storage capacitor preparation process with normal pixels and not participating in real-time frame display), and writes the reference voltage corresponding to the same standard gray level as the normal pixels (such as 5V corresponding to 100% gray level, and this voltage is the same as the initial voltage signal written by the column data sub-circuit of the driving control unit to the normal pixels) to the storage capacitor of the redundant pixels through the column data sub-circuit of the driving control unit; the hardware calibration module collects the actual emission brightness during the emission of the redundant pixels (i.e., during the TFT conduction stage when the row scanning sub-circuit selects the TFT of the redundant pixel) through an array of optical sensors integrated above the color filter layer and corresponding to the pixel units one by one (one sensor is arranged directly above each redundant pixel); the hardware calibration module calculates the deviation value between the actual brightness and the preset standard brightness (such as 400 cd / m², which is the target brightness of normal pixels in a standard environment), classifies and stores it according to the interval of the real-time decay rate of the storage capacitor calculated by the voltage sampling sub-module (such as 5 - 10 mV / s, 10 - 15 mV / s), and after establishing a "decay rate - brightness deviation" mapping table, transmits it to the feed-forward compensation module (for correcting the "decay trend - voltage increment" mapping relationship) and the feedback compensation module (for correcting the "decay trend - compensation pulse amplitude" parameter) of the dynamic compensation unit through the system bus.

[0060] The hardware calibration module of the self-learning calibration unit is a basic component for the system to achieve long-term performance optimization. Its core function is to establish the correlation between the decay rate of the storage capacitor and the display brightness deviation through the periodic calibration of redundant pixels, providing experimental data support for the parameter optimization of the dynamic compensation unit. The working process of this module strictly follows the standardized process of "trigger - execution - data processing - transmission", and the specific implementation steps are as follows: The trigger condition of the hardware calibration module is to start once every 100 hours. This cycle is set based on the typical device aging time scale of the LED display matrix (such as the performance drift of the dielectric material of the storage capacitor usually stabilizes after hundreds of hours), which can not only capture the device characteristic changes in time but also avoid the interference of frequent calibration on normal display. The calibration object is selected as 2 rows × 2 columns of redundant pixels reserved at the edge of the display matrix. These pixels share the TFT driving circuit process (such as the same TFT channel length and gate insulation layer thickness) and the storage capacitor preparation process (such as the same dielectric material deposition parameters and electrode patterning accuracy) with normal display pixels and do not participate in the current frame's real-time display (shielding the TFT conduction through the selection signal of the row scanning sub-circuit), so as to ensure that the calibration process does not affect the normal picture display.

[0061] The hardware calibration module writes a reference voltage to the storage capacitor of the redundant pixel through the column data sub-circuit of the drive control unit. The writing operation is consistent with the voltage writing process of normal display pixels: the digital-to-analog converter of the column data sub-circuit converts the standard gray value (such as 100% gray) into a corresponding analog voltage signal (such as 5V), and transmits it to the storage capacitor of the redundant pixel through the column data line. Since the storage capacitors of redundant pixels and normal pixels have exactly the same process, and the written voltage is the same as the initial voltage signal of normal pixels (such as 5V corresponding to 100% gray), the process of writing the reference voltage can truly simulate the storage capacitor voltage loading scenario under normal working conditions.

[0062] The brightness acquisition of the array-type optical sensor is a key link in calibration. The sensor is integrated above the color filter layer (on the light-emitting side of the LED diode), and a sensor is set directly above each redundant pixel (corresponding one-to-one with the pixel unit) to achieve the acquisition of the luminous brightness of only the corresponding pixel and avoid crosstalk. The brightness acquisition is carried out during the luminous period of the redundant pixel: the row scanning sub-circuit outputs a gating signal to activate the TFT gate of the redundant pixel. After the TFT is turned on, the storage capacitor maintains the gate voltage, and the LED diode emits light due to current conduction; the sensor continuously acquires the light intensity signal during the entire period when the TFT is turned on (that is, during the high level period of the row gating signal), and converts it into a digital brightness value through the analog-to-digital conversion circuit (such as 420 cd / m²).

[0063] The hardware calibration module calculates the deviation value based on the acquired brightness data. The preset standard brightness (such as 400 cd / m²) is the target brightness of normal pixels in the standard environment (25°C, no aging), which is determined by calibrating the all-white screen during system initialization. The module calculates the absolute deviation between the actual brightness (such as 420 cd / m²) and the standard brightness (|420 cd / m² - 400 cd / m²| = 20 cd / m²), and according to the real-time attenuation rate of the storage capacitor (such as 8 mV / µs) synchronously calculated by the voltage sampling sub-module, classifies the deviation value into the corresponding attenuation rate interval (such as 5 - 10 mV / µs). This classification and storage process establishes a mapping relationship between "attenuation rate - brightness deviation" (such as the average deviation of 20 cd / m² corresponding to the 5 - 10 mV / µs interval, and the average deviation of 35 cd / m² corresponding to the 10 - 15 mV / µs interval), intuitively reflecting the influence degree of the storage capacitor attenuation rate on the display brightness.

[0064] Finally, the "attenuation rate - luminance deviation" mapping table is transmitted to the feedforward compensation module and the feedback compensation module of the dynamic compensation unit through the system bus. The feedforward compensation module uses this table to correct the "attenuation trend - voltage increment" mapping relationship (for example, the original "rapid attenuation" corresponds to +1.2V. If the luminance deviation in the range of 5 - 10 mV / μs is 20 cd / m² exceeding the threshold, it is adjusted to +1.5V to increase the voltage redundancy); the feedback compensation module uses this table to correct the "attenuation trend - compensation pulse amplitude" parameter (for example, the original "rapid attenuation" corresponds to 0.3V. If the luminance deviation in the range of 10 - 15 mV / μs is 35 cd / m², it is adjusted to 0.4V to enhance the compensation effect).

[0065] The hardware calibration module correlates the physical attenuation characteristics of the storage capacitor with the actual performance of the display luminance through the periodic calibration of redundant pixels, providing real experimental data support for the parameter optimization of the dynamic compensation unit. This design enables the system to adaptively adjust the control strategy with device aging and environmental changes, which is the key technical guarantee for achieving long-term luminance uniformity of the display screen.

[0066] Specifically, in the intelligent control system for micro-LED displays of the present invention, the self-learning calibration unit includes a model calibration module; the model calibration module obtains the "attenuation rate and luminance deviation" mapping table established by the hardware calibration module. For each attenuation rate interval in the table (such as 5 - 10 mV / s), it extracts the corresponding temperature value (the regional temperature average value from the temperature sensing sub-module of the state prediction unit), the usage duration (the total pixel working duration from the usage duration recording sub-module), the attenuation rate (the real-time attenuation rate from the voltage sampling sub-module), and the luminance deviation value (the calculation result from the hardware calibration module), and compares the luminance deviation with a preset threshold (such as ±20 cd / m², which is the maximum luminance fluctuation allowed for normal display); if the luminance deviation exceeds the threshold, it corrects the "attenuation trend and voltage increment" mapping relationship of the feedforward compensation module of the dynamic compensation unit (such as the original rapid attenuation increasing by 1.2V → adjusted to 1.5V) or the "attenuation trend and compensation pulse amplitude" parameter of the feedback compensation module (such as the original 0.3V → 0.4V); simultaneously, it adds the extracted temperature, usage duration, attenuation rate, and luminance deviation data to the historical database of the state prediction unit, updates the boundary values of the three-dimensional index of "temperature interval - usage duration interval - attenuation rate difference" based on the distribution characteristics of the new data (such as the original 20 - 30 °C → adjusted to 18 - 32 °C), and re-statistics the typical attenuation trends under each index (such as the original rapid attenuation threshold > 10 mV / s → adjusted to > 8 mV / s, based on the attenuation rate distribution of more than 80% of the samples in the new data), so as to optimize the prediction accuracy of the pattern matching algorithm of the trend prediction module.

[0067] The model calibration module of the self-learning calibration unit is the core decision-making component for the system to achieve adaptive optimization. By analyzing the experimental data provided by the hardware calibration module, it dynamically corrects the dynamic compensation strategy and updates the trend prediction model to achieve the control accuracy of the system's long-term operation. The workflow of this module closely revolves around "data extraction - deviation judgment - parameter correction - database update", and the specific implementation steps are as follows: The primary operation of the model calibration module is to obtain the "attenuation rate - brightness deviation" mapping table established by the hardware calibration module. This table is a structured data file, and each row corresponds to an attenuation rate interval (such as 5 - 10 mV / μs, 10 - 15 mV / μs), and is associated with the temperature value within this interval (the regional temperature average from the temperature sensing sub-module of the state prediction unit, such as 25°C), the usage duration (the total pixel working duration from the usage duration recording sub-module, such as 800 hours), the attenuation rate (the real-time attenuation rate from the voltage sampling sub-module, such as 8 mV / μs), and the brightness deviation value (the calculation result from the hardware calibration module, such as +20 cd / m²). The data in the table is transmitted from the hardware calibration module to the storage unit of the model calibration module through the system bus, serving as the basic input for subsequent analysis.

[0068] For each attenuation rate interval in the table, the model calibration module extracts the associated multi-dimensional data. For example, for the "5 - 10 mV / μs" interval, the extracted temperature value is 25°C (corresponding to the regional temperature average of the temperature sensing sub-module), the usage duration is 800 hours (corresponding to the total pixel working duration of the usage duration recording sub-module), the attenuation rate is 8 mV / μs (corresponding to the real-time attenuation rate of the voltage sampling sub-module), and the brightness deviation value is +20 cd / m² (corresponding to the calculation result of the hardware calibration module). These data completely record the operating environment, device status, and display effect under this attenuation rate interval, providing multi-source basis for deviation judgment.

[0069] The model calibration module needs to judge whether the brightness deviation of each attenuation rate interval exceeds the acceptable range. The preset threshold (such as ±20 cd / m²) is the maximum brightness fluctuation allowed for normal display, which is determined by the full-white screen uniformity test during system initialization (such as 95% of the normal pixel brightness deviations are within ±20 cd / m²). The module traverses each interval of the mapping table and compares the brightness deviation of this interval with the threshold: if the deviation value (such as +20 cd / m²) does not exceed the threshold (±20 cd / m²), it is marked as "normal"; if the deviation value (such as +25 cd / m²) exceeds the threshold, it is marked as "abnormal", triggering the parameter correction process.

[0070] For the attenuation rate intervals marked as "abnormal", the model calibration module corrects the control parameters of the dynamic compensation unit. Taking the "attenuation trend - voltage increment" mapping relationship of the feedforward compensation module as an example: if the brightness deviation in the interval of "5 - 10 mV / μs" is +25 cd / m² (exceeding the threshold of +20 cd / m²), it indicates that the current voltage increment (such as the original +1.2 V) is insufficient to compensate for the brightness deficiency caused by attenuation. The module adjusts the voltage increment corresponding to this trend to +1.5 V (increasing the redundancy by 0.3 V) to enhance the compensation effect. Similarly, for the "attenuation trend - compensation pulse amplitude" parameter of the feedback compensation module, if the brightness deviation in the interval of "10 - 15 mV / μs" is +35 cd / m² (exceeding the threshold), the pulse amplitude corresponding to this trend is adjusted from 0.3 V to 0.4 V (increasing the compensation amount by 0.1 V) to improve the real-time correction ability.

[0071] The model calibration module synchronously adds the extracted multi-dimensional data to the historical database of the state prediction unit. The new data includes temperature values (such as 25 °C), usage duration (such as 800 hours), attenuation rate (such as 8 mV / μs), and brightness deviation values (such as +25 cd / m²). These data are classified and stored according to the three-dimensional index of "temperature interval - usage duration interval - attenuation rate difference". Based on the distribution characteristics of the new data (such as the minimum temperature value of 18 °C, the maximum temperature value of 32 °C, and the concentrated interval of the attenuation rate being 6 - 10 mV / μs), the module adjusts the boundary values of the three-dimensional index (such as expanding the original interval of "20 - 30 °C" to "18 - 32 °C" to cover a wider temperature range). At the same time, the typical attenuation trends under each index are re-counted: if more than 80% of the samples in the new data show an attenuation rate > 8 mV / μs under the index of "18 - 32 °C - 500 - 1000 hours - ΔR = 5 - 10 mV / μs", the typical trend of this index is updated from "stable attenuation" to "rapid attenuation", and the trend threshold is adjusted from the original > 10 mV / μs to > 8 mV / μs (based on the concentrated distribution of the new samples).

[0072] Through the above process, the model calibration module converts the experimental data of hardware calibration into optimization instructions for control parameters and update basis for the trend prediction model, enabling the system to dynamically adjust the control strategy with device aging and environmental changes, and continuously maintaining the precise control ability over the voltage attenuation of the storage capacitor. This design realizes a closed-loop self-learning mechanism of "data collection - analysis - optimization", which is the core technical guarantee for the long-term reliability and display uniformity of the micro-LED display intelligent control system.

[0073] Specifically, in the micro-LED display intelligent control system of the present invention, the auxiliary reference capacitor is integrated in the edge calibration area of the non-display area of the display matrix (physically isolated from the pixel units in the display area and not connected to the LED diodes), and is prepared using the same batch of dielectric materials (such as silicon nitride thin films of the same thickness) and the same process parameters (such as the same magnetron sputtering deposition temperature and electrode patterning lithography accuracy) as the storage capacitor; both ends of the auxiliary reference capacitor are connected in parallel with a micro-current monitoring circuit (the same as the sampling circuit of the voltage sampling sub-module), which real-time collects its leakage current and calculates the attenuation rate (attenuation rate = leakage current × capacitor plate area / dielectric constant). This attenuation rate data is transmitted to the attenuation rate difference calculation unit of the trend prediction module of the state prediction unit through an independent data bus; since there is no LED diode load in the calibration area (only including the capacitor structure), the attenuation rate data of the auxiliary reference capacitor can be used as a reference value without display load interference to calculate the difference between the real-time attenuation rate of the storage capacitor and the real-time attenuation rate of the auxiliary reference capacitor (ΔR = real-time attenuation rate of the storage capacitor - real-time attenuation rate of the auxiliary reference capacitor).

[0074] The auxiliary reference capacitor is a key component in the micro-LED display intelligent control system for obtaining the intrinsic attenuation characteristics of the storage capacitor. Its design and working process closely revolve around the three core objectives of "no load interference", "process consistency", and "data comparability". The specific implementation details are as follows: The physical layout of the auxiliary reference capacitor is the basis for achieving no load interference. It is integrated in the edge calibration area of the non-display area of the display matrix (such as the non-display rows at the upper and lower edges of the matrix), and is physically isolated from the pixel units in the display area through an insulating isolation layer (such as a silicon dioxide isolation wall), and is not connected to the LED diodes (only the capacitor structure is retained). This layout avoids the influence of the display load (such as the current consumption of the LED diodes) on the capacitor attenuation characteristics, and enables the attenuation of the auxiliary reference capacitor to be caused only by the leakage current of the dielectric material itself, thus providing a pure intrinsic attenuation reference for the attenuation analysis of the storage capacitor.

[0075] The preparation process of the auxiliary reference capacitor is highly consistent with that of the storage capacitor, which is the key to ensuring the comparability of the intrinsic attenuation characteristics. The two use the same batch of dielectric materials (such as silicon nitride thin films of the same batch, with a thickness deviation ≤ 0.5%) and exactly the same preparation process parameters: the magnetron sputtering deposition temperature is strictly controlled at 300 ± 5 °C (the same as the storage capacitor) to achieve the density and uniformity of the dielectric layer; the electrode patterning uses the same set of lithography masks (line width accuracy ± 0.1 μm) to ensure that the area and spacing of the capacitor plates are exactly the same as those of the storage capacitor. This process consistency makes the intrinsic attenuation characteristics (such as leakage current density and dielectric constant) of the auxiliary reference capacitor and the storage capacitor theoretically exactly the same, and the difference is only introduced by the subsequent load, providing a reliable reference for the subsequent difference calculation.

[0076] The decay rate monitoring of the auxiliary reference capacitor is achieved through a shunt micro-current monitoring circuit. This circuit uses the same model as the sampling circuit of the voltage sampling sub-module of the state prediction unit (such as the AD8628 high-input impedance operational amplifier from Analog Devices, Inc.), ensuring the same sampling principle and accuracy: the circuit converts the capacitor leakage current into a voltage signal through a high-input impedance transimpedance amplifier (such as 1 μA leakage current corresponding to 1 mV voltage output). After filtering out high-frequency noise by a low-pass filter (cutoff frequency 1 kHz), it is converted into a digital signal by an analog-to-digital converter (ADC). The output of the micro-current monitoring circuit is transmitted to the trend prediction module of the state prediction unit through an independent data bus (physically isolated from the display data bus) to avoid interference from the high-frequency noise of the display data on the decay rate calculation.

[0077] The calculation of the decay rate is based on the intrinsic physical properties of the capacitor. Although the specific formula (decay rate = leakage current × capacitor plate area / dielectric constant) is not directly output, the monitoring circuit can indirectly reflect the voltage decay rate of the capacitor through the real-time collected leakage current value (such as 0.5 μA), the known capacitor plate area (such as 100 μm²), and the dielectric constant (such as the dielectric constant of silicon nitride ε = 7) (the larger the leakage current, the faster the voltage decays per unit time). This rate data is used as a reference value without load interference and is transmitted to the decay rate difference calculation unit of the trend prediction module.

[0078] The calculation of the decay rate difference is the core function of the auxiliary reference capacitor. The difference calculation unit receives the real-time decay rate of the storage capacitor (such as 2.5 mV / μs, including the influence of the load) and the reference decay rate of the auxiliary reference capacitor (such as 0.5 mV / μs, only intrinsic decay), and calculates the difference between the two (ΔR = 2.5 mV / μs - 0.5 mV / μs = 2.0 mV / μs). This difference directly quantifies the additional decay amount of the storage capacitor due to connecting the LED diode load and is a key parameter for distinguishing intrinsic decay and load influence. Based on this difference, the trend prediction module combines temperature and usage duration data to judge the decay trend of the storage capacitor (such as ΔR > 1.5 mV / μs marked as "rapid decay"), providing accurate control basis for the dynamic compensation unit.

[0079] Through the above design, on the premise of ensuring the same intrinsic decay characteristics as the storage capacitor, the auxiliary reference capacitor provides reliable reference decay data for the system through an environment without load interference and a highly consistent monitoring circuit. This design enables the system to accurately distinguish the intrinsic factors and load factors of the storage capacitor decay, which is a prerequisite for realizing dynamic compensation and self-learning calibration and is also the core technical support for the micro-LED display intelligent control system to have high-precision control capabilities.

[0080] Specifically, in the micro-LED display intelligent control system of the present invention, the temperature sensing sub-module is a thin-film thermocouple integrated in the pixel gap area of the TFT substrate (one is set at the center position of each 2×2 pixel unit, and a total of m×n thermocouples are set, where m and n are the number of rows and columns of the display matrix). Its hot end is in contact with the TFT substrate of the pixel unit (for sensing the Joule heat when the pixel is working), and the cold end is connected to the constant-temperature reference point at the edge of the substrate (for eliminating the influence of ambient temperature fluctuations); the thin-film thermocouple collects the temperature values of each distribution point in real time (such as the temperature T1 at point 1 and the temperature T2 at point 2). After converting the analog temperature signal into a digital signal through the analog-to-digital conversion circuit, the multi-dimensional monitoring module of the state prediction unit eliminates outliers (such as extreme values exceeding the temperature tolerance upper limit of the substrate material), and then calculates the average temperature of the remaining valid data (T_avg=(T1+T2+…+Tk) / k, k≤m×n), which is output as the regional average temperature of the display matrix to the temperature interval matching unit of the trend prediction module; the temperature interval matching unit compares T_avg with the boundary values of each temperature interval in the historical database (such as 20-30°C, 30-40°C) to determine the real-time temperature interval (such as 25°C belongs to the 20-30°C interval).

[0081] The temperature sensing sub-module is the core component for realizing temperature sensing in the micro-LED display intelligent control system. It collects the regional temperature data of the display matrix through a distributed thin-film thermocouple array, and provides environmental temperature parameters for the trend prediction module after data processing. The design and working process of this module closely revolve around the three goals of "accurate sensing", "reliable data", and "environmental interference resistance". The specific implementation details are as follows: The physical layout of the thin-film thermocouple is the basis for realizing accurate temperature sensing. The module integrates the thin-film thermocouple in the pixel gap area of the TFT substrate (that is, the non-display area between adjacent pixel units), and adopts a distribution strategy of setting one at the center position of each 2×2 pixel unit (for example, when the display matrix is 100×100 pixels, a total of 50×50 thermocouples are set). This uniform distribution enables the temperature acquisition points to cover the entire display area, avoiding measurement errors caused by local temperature deviations. The hot end (sensitive end) of the thermocouple is directly in contact with the TFT substrate through a lithography process, and the contact area covers the TFT active layer of the pixel unit (the temperature of this area is the highest due to the Joule heat when the TFT is turned on), so as to realize the sensing of the real temperature when the pixel is working; the cold end (reference end) is connected to the constant-temperature reference point at the edge of the substrate through metal wiring (such as a constant-temperature area maintained at 25±0.5°C by a thermoelectric cooler), eliminating the interference of ambient temperature fluctuations on the measurement results, and ensuring that the thermocouple output only reflects the difference between the pixel working temperature and the reference temperature.

[0082] The acquisition and conversion of temperature signals are the prerequisite for data processing. The thin-film thermocouple uses nickel-chromium-nickel-silicon (Type K) thermoelectric materials, and its output voltage is proportional to the temperature difference (thermoelectric power rate is about 41 μV / °C). Each thermocouple outputs an analog voltage signal in real time (for example, when the hot-end temperature is 30°C and the cold-end temperature is 25°C, the output is 205 μV). After being amplified by a low-noise differential amplifier (such as AD8421 from ADI), it is transmitted to the analog-to-digital conversion circuit (ADC, resolution 12 bits, sampling rate 1 kHz). The ADC converts the analog voltage signal into a digital temperature value (for example, 205 μV corresponds to a temperature difference of 5°C, and with a cold-end temperature of 25°C, the actual temperature is 30°C), and is transmitted to the multi-dimensional monitoring module of the state prediction unit through the SPI bus.

[0083] Data preprocessing is a crucial step in enhancing the reliability of temperature data. The multi-dimensional monitoring module eliminates outliers from the acquired digital temperature values: First, set the upper temperature limit of the substrate material (for example, the upper temperature limit of the polyimide material commonly used in TFT substrates is 150°C). If the measured temperature of a certain thermocouple exceeds this threshold (such as 160°C), it is determined as an outlier (possibly caused by sensor failure or electromagnetic interference); Second, calculate the mean and standard deviation of the temperatures of all thermocouples (for example, mean 30°C, standard deviation 2°C), and eliminate extreme values that deviate more than 3 times the standard deviation from the mean (such as data outside 30°C ± 6°C). After eliminating outliers, the module calculates the arithmetic mean (T_avg=(T1+T2+…+Tk) / k) of the remaining valid data (k≤m×n, m and n are the number of rows and columns of the display matrix) as the regional temperature average of the current display matrix (such as 30.5°C).

[0084] Temperature range matching is a bridge connecting real-time data and historical data. The temperature range matching unit of the trend prediction module receives the regional temperature average (such as 30.5°C) output by the multi-dimensional monitoring module and compares it one by one with the predefined temperature ranges (such as 20 - 30°C, 30 - 40°C, 40 - 50°C) in the historical database. The matching process is achieved through numerical comparison: If T_avg = 30.5°C, it falls into the "30 - 40°C" range; if T_avg = 25°C, it falls into the "20 - 30°C" range. The temperature ranges in the historical database are divided based on the typical operating temperature range of the display matrix (such as consumer electronics displays usually operate at 20 - 50°C), and each range corresponds to the typical temperature characteristics of the storage capacitor dielectric material (such as when the temperature increases by 10°C, the leakage current increases by 20%).

[0085] Through the above process, the temperature sensing sub-module realizes the full-link processing from temperature perception, signal conversion, data cleaning to interval matching, providing reliable parameters reflecting the overall temperature environment of the display matrix for the trend prediction module. This design enables the system to accurately evaluate the impact of temperature on the decay characteristics of storage capacitors, which is a key input for the multi-dimensional matching of the trend prediction module and an important basis for the dynamic compensation unit to formulate precise control strategies.

[0086] Through the hierarchical cooperation of the drive control unit, state prediction unit, dynamic compensation unit, and self-learning calibration unit, the present invention constructs a closed-loop control link of "monitoring - prediction - compensation - calibration" to specifically solve the problem of uneven brightness in the gray area of the display screen caused by the coupling failure between the voltage maintenance stability of the storage capacitor and the consistency of dynamic brightness regulation. As the basic signal transmission module, the drive control unit provides the row selection duration, the initial voltage amplitude, and the reference decay rate of the auxiliary reference capacitor through the row scanning sub-circuit, column data sub-circuit, and dual-capacitor module respectively, providing the original data support for subsequent monitoring and prediction; the state prediction unit, through the multi-dimensional monitoring module (including the voltage sampling sub-module, temperature sensing sub-module, and usage duration recording sub-module), real-time collects the residual voltage of the storage capacitor, the regional temperature, and the total pixel working duration, calculates the decay rate difference in combination with the reference data of the drive control unit, and predicts the decay trend using the three-dimensional index and pattern matching algorithm of the historical database, providing a forward-looking basis for dynamic compensation.

[0087] Based on the predicted trend, the dynamic compensation unit cancels the uncertainty of the storage capacitor decay in two stages: feedforward and feedback. The feedforward compensation module receives the trend prediction result, calls the "decay trend - voltage increment" and "decay trend - gating timing offset" mapping relationships optimized by self-learning calibration, and adjusts the initial voltage amplitude of the column data sub-circuit and the gating timing of the row scanning sub-circuit in advance. By increasing the voltage redundancy and extending the effective conduction duration, it actively copes with the decay risk; the feedback compensation module monitors the storage capacitor voltage in real time at a frequency of 10 kHz during pixel emission, calculates the real-time decay rate. If it exceeds the trend prediction threshold, a narrow pulse compensation voltage is injected during the adjacent row blanking period to correct the residual deviation of the feedforward adjustment, and the decay data is synchronously fed back to verify the prediction accuracy.

[0088] The self-learning calibration unit realizes the long-term optimization of system performance through hardware calibration and model calibration. The hardware calibration module calls redundant pixels every 100 hours, writes the standard gray reference voltage and collects the actual brightness, and establishes a "decay rate - brightness deviation" mapping table; the model calibration module extracts multi-dimensional data based on this table, compares the brightness deviation with the preset threshold, corrects the mapping relationship and parameters of the dynamic compensation unit, synchronously updates the three-dimensional index and typical trend of the historical database of the state prediction unit, and optimizes the prediction accuracy of the pattern matching algorithm. Through this closed-loop mechanism, the system can adapt to the performance drift of the storage capacitor caused by environmental temperature changes and device aging, continuously maintain the matching of the TFT conduction time and the voltage holding period, and finally achieve the brightness uniformity of the same gray area of the display screen.

Claims

1. A micro-LED display intelligent control system, characterized in that: A drive control unit, a state prediction unit, a dynamic compensation unit, and a self-learning calibration unit; the drive control unit outputs the row strobe duration through the row scanning sub-circuit, outputs the initial voltage amplitude of the storage capacitor through the column data sub-circuit, and outputs the real-time attenuation rate to the state prediction unit through the auxiliary reference capacitor; The state prediction unit includes a multi-dimensional monitoring module, which collects the residual voltage of the storage capacitor through the voltage sampling sub-module and calculates the real-time attenuation rate, collects the average regional temperature through the temperature sensing sub-module, accumulates the total working duration of the pixels through the usage duration recording sub-module, combines the initial voltage, row strobe duration, and auxiliary reference capacitor attenuation rate output by the drive control unit, and uses the historical database and pattern matching algorithm in the trend prediction module to predict the attenuation trend of the storage capacitor, and transmits the prediction result to the dynamic compensation unit; The dynamic compensation unit adjusts the voltage amplitude of the column data sub-circuit and the strobe timing of the row scanning sub-circuit through the feed-forward compensation module, and monitors the real-time attenuation rate of the storage capacitor during pixel emission through the feedback compensation module and feeds it back to the trend prediction module of the state prediction unit for model verification; The self-learning calibration unit calls the redundant pixels at the edge of the display matrix through the hardware calibration module to generate a mapping table of attenuation rate and brightness deviation. Based on the mapping table of attenuation rate and brightness deviation, the model calibration module updates the prediction parameters of the historical database of the state prediction unit, and synchronously corrects the compensation parameters of the feed-forward module and the feedback module of the dynamic compensation unit.

2. The micro-LED display intelligent control system according to claim 1, wherein: The drive control unit includes a row scanning sub-circuit, a column data sub-circuit, and a dual-capacitor module; The row scanning sub-circuit generates a row strobe signal by the timing controller, and outputs a high-level signal row by row in the order of row addresses, activating the TFT gate of the corresponding row pixel unit to enter the conduction state; the column data sub-circuit generates a column data signal by the data driver, and only writes the initial voltage signal corresponding to the image gray level to the storage capacitors of all pixel units in that row during the period when the corresponding row strobe signal is high; The dual-capacitor module includes a storage capacitor connected in parallel between the TFT gate and the ground and an auxiliary reference capacitor integrated in the non-display area of the display matrix; The row scanning sub-circuit outputs the row strobe duration to the voltage sampling sub-module of the state prediction unit, the column data sub-circuit outputs the initial voltage amplitude of the storage capacitor to the voltage sampling sub-module of the state prediction unit, and the auxiliary reference capacitor outputs the real-time attenuation rate to the trend prediction module of the state prediction unit through the micro-current monitoring circuit.

3. The micro-LED display intelligent control system according to claim 2, characterized in that: The state prediction unit includes a multi-dimensional monitoring module; the multi-dimensional monitoring module includes a voltage sampling sub-module, a temperature sensing sub-module, and a usage duration recording sub-module; at the end of each row strobe period, the voltage sampling sub-module collects the residual voltage value of the storage capacitor through the micro-current sampling circuit connected in parallel at both ends of the storage capacitor, and calculates the attenuation rate of the storage capacitor within the real-time period based on the initial voltage amplitude output by the column data sub-circuit of the drive control unit and the row strobe duration output by the row scanning sub-circuit; The temperature sensing sub-module integrates thin-film thermocouples in the pixel gap region of the TFT substrate, and collects the temperature values of each distribution point in real time. After converting the analog temperature signal into a digital signal through the analog-to-digital conversion circuit, it calculates the average temperature value of all distribution points as the regional average temperature; The usage duration recording sub-module configures a non-volatile counter for each pixel unit, and only accumulates and records the conduction time during the TFT conduction period. When the system starts, it reads and updates the total working duration of the pixel unit; The multi-dimensional monitoring module transmits the real-time attenuation rate of the storage capacitor calculated by the voltage sampling sub-module to the attenuation rate difference calculation unit of the trend prediction module, transmits the regional average temperature output by the temperature sensing sub-module to the temperature range matching unit of the trend prediction module, and transmits the total working duration of the pixel recorded by the usage duration recording sub-module to the usage duration range matching unit of the trend prediction module.

4. The micro-LED display intelligent control system according to claim 3, characterized in that: The state prediction unit includes a trend prediction module; the trend prediction module is provided with a historical database, and the historical database constructs a three-dimensional index according to temperature ranges, usage duration ranges, and attenuation rate differences, and stores typical attenuation trends under different indexes; The attenuation rate difference calculation unit of the trend prediction module receives the real-time attenuation rate of the storage capacitor transmitted by the multi-dimensional monitoring module and the real-time attenuation rate output by the auxiliary reference capacitor of the drive control unit, and calculates the difference between the two; The temperature range matching unit receives the regional average temperature transmitted by the multi-dimensional monitoring module and matches the corresponding temperature range in the historical database; The usage duration range matching unit receives the total working duration of the pixel transmitted by the multi-dimensional monitoring module and matches the corresponding usage duration range in the historical database; The trend prediction module uses a pattern matching algorithm to input the matched temperature range, usage duration range, and attenuation rate difference ΔR into the historical database for matching degree calculation, selects the typical attenuation trend corresponding to the index with the highest matching degree, and transmits it to the feed-forward compensation module of the dynamic compensation unit as the attenuation trend prediction result of the storage capacitor in the next frame period.

5. The micro-LED display intelligent control system according to claim 4, characterized in that: The dynamic compensation unit includes a feed-forward compensation module; The feed-forward compensation module receives the attenuation trend prediction result of the next frame period output by the trend prediction module, and adjusts the initial voltage amplitude of the column data sub-circuit based on the attenuation trend and voltage increment mapping relationship updated by the model calibration module of the self-learning calibration unit; Synchronously based on the attenuation trend and strobe timing offset mapping relationship updated by the model calibration module, adjust the strobe timing of the row scanning sub-circuit; The feed-forward compensation module transmits the adjusted voltage amplitude parameter to the digital-to-analog converter of the column data sub-circuit, and transmits the adjusted strobe timing parameter to the timing controller of the row scanning sub-circuit, and the drive control unit performs the updated voltage writing and row strobe operations.

6. The micro-LED display intelligent control system according to claim 5, wherein: The dynamic compensation unit includes a feedback compensation module; During the pixel emission period, the feedback compensation module monitors the voltage value across the storage capacitor in real time at a frequency of 10 kHz through the voltage sampling sub-module, and calculates the real-time attenuation rate of the storage capacitor based on the difference between the initial write voltage adjusted by the feed-forward compensation module and the real-time monitored voltage, and the monitoring time interval; The feedback compensation module obtains the typical threshold corresponding to the decay trend in the historical database of the trend prediction module. If the real-time decay rate exceeds this threshold, a narrow pulse compensation voltage is injected into the storage capacitor during the horizontal blanking period between the strobe signals of adjacent rows through the column data sub-circuit. The feedback compensation module transmits the real-time calculated decay rate of the storage capacitor to the trend prediction module of the state prediction unit, which is used to compare with the predicted decay trend in the historical database, verify the prediction accuracy of the pattern matching algorithm, and trigger the parameter update of the model calibration module.

7. The micro-LED display intelligent control system according to claim 6, wherein: The self-learning calibration unit includes a hardware calibration module. The hardware calibration module calls 2 rows × 2 columns of redundant pixels reserved at the edge of the display matrix every 100 hours, and writes the reference voltage corresponding to the standard gray level same as that of the normal pixels into the storage capacitors of the redundant pixels through the column data sub-circuit of the drive control unit. The hardware calibration module collects the actual emission brightness of the redundant pixels during the emission period of the redundant pixels through an array-type optical sensor integrated above the color filter layer and corresponding to the pixel unit. The hardware calibration module calculates the deviation value between the actual brightness and the preset standard brightness, classifies and stores it according to the interval of the real-time decay rate of the storage capacitor calculated by the voltage sampling sub-module. After establishing the decay rate and brightness deviation mapping table, it is transmitted to the feed-forward compensation module and the feedback compensation module of the dynamic compensation unit through the system bus.

8. The micro-LED display intelligent control system according to claim 7, characterized in that: The self-learning calibration unit includes a model calibration module. The model calibration module obtains the decay rate and brightness deviation mapping table established by the hardware calibration module. For each decay rate interval in the table, it extracts the corresponding temperature value, usage duration, decay rate, and brightness deviation value in this interval, and compares the brightness deviation with the preset threshold. If the brightness deviation exceeds the threshold, the decay trend and voltage increment mapping relationship of the feed-forward compensation module of the dynamic compensation unit or the decay trend and compensation pulse amplitude parameter of the feedback compensation module are corrected. Synchronously add the extracted temperature, usage duration, decay rate, and brightness deviation data to the historical database of the state prediction unit, update the boundary values of the three-dimensional index of the temperature interval, usage duration interval, and decay rate difference based on the distribution characteristics of the new data, and re-statistics the typical decay trend under each index to optimize the prediction accuracy of the pattern matching algorithm of the trend prediction module.

9. The micro-LED display intelligent control system according to claim 2, characterized in that: The auxiliary reference capacitor is integrated in the edge calibration area of the non-display area of the display matrix, and is prepared with the same batch of dielectric materials and the same process parameters as the storage capacitor. A micro-current monitoring circuit is connected in parallel at both ends of the auxiliary reference capacitor to collect its leakage current in real time and calculate the decay rate. This decay rate data is transmitted to the decay rate difference calculation unit of the trend prediction module of the state prediction unit through an independent data bus. Since there is no LED diode load in the calibration area, the decay rate data of the auxiliary reference capacitor can be used as a reference value without display load interference, and is used to calculate the difference between the real-time decay rate of the storage capacitor and the real-time decay rate of the auxiliary reference capacitor.

10. The micro-LED display intelligent control system according to claim 3, wherein: The temperature sensing sub-module is a thin-film thermocouple integrated in the pixel gap area of the TFT substrate. Its hot end is in contact with the TFT substrate of the pixel unit, and its cold end is connected to the constant temperature reference point at the edge of the substrate. The thin-film thermocouple collects the temperature values of each distribution point in real time. After converting the analog temperature signal into a digital signal through an analog-to-digital conversion circuit, the multi-dimensional monitoring module of the state prediction unit eliminates outliers, and then calculates the average temperature of the remaining valid data, which is output as the average temperature of the region of the display matrix to the temperature interval matching unit of the trend prediction module; the temperature interval matching unit compares T_avg with the boundary values of each temperature interval in the historical database to determine the temperature interval to which it belongs in real time.

Images