Energy-saving control method and system for landscape LED lamp array based on zoning load balancing

Abstract

The application discloses a landscape LED lamp array energy-saving control method and system based on partition load balancing and belongs to the field of novel lighting, and comprises the following steps: after partitioning the lamp array, aligning three types of data every period to generate a brightness boundary, recursively re-negotiating if the lower boundary exceeds the upper boundary, distributing time sequence offset according to current margin to make the brightness change smoothly and averagely remain unchanged if the total power peak value exceeds the limit, or directly outputting a typical brightness. The application precisely aligns multi-source asynchronous data, eliminates control deviation caused by sampling period conflict and communication delay, solves the problems of brightness jump, branch overcurrent and unstable energy-saving indicators, and then distributes time sequence offset according to current margin sorting and recursively re-negotiates the boundary, so that the power peak value is dispersed and constraint convergence is realized under complex topology. On the premise of guaranteeing visual smoothness and electrical safety, the application reduces hardware cost, improves system robustness and stably achieves preset energy-saving indicators.

Description

Technical Field

[0001] This invention relates to the field of novel lighting technology, and in particular to an energy-saving control method and system for landscape LED light arrays based on zoned load balancing. Background Technology

[0002] In landscape LED light array control, existing systems typically obtain the target brightness of each zone based on program instructions and drive the corresponding LED groups to achieve brightness output through pulse width modulation (PWM) signals. To improve the current stability and constant current drive effect during PWM dimming, the relevant drive circuits usually introduce sampling and comparison control mechanisms within multiple consecutive switching cycles of the power switching transistors. By generating effective switching control signals in the first few switching cycles, the sampling window and the drive cycle are synchronized, thereby improving the stability and smoothness of the dimming process.

[0003] Based on the above-mentioned drive control, in order to achieve power supply safety and energy consumption optimization, a real-time monitoring mechanism based on current, voltage and branch remaining capacity is further introduced, and the power demand for the next control cycle is predicted by combining historical operating data, thereby providing forward-looking constraints on power scheduling in the time dimension.

[0004] As the scale of LED arrays increases, a single global control method becomes increasingly inadequate to simultaneously address local temperature rise differences and luminous uniformity. Therefore, existing technologies further introduce a dynamic zoning control approach: by performing topology analysis on the control nodes of parallel branches, the initial impedance of each branch is calculated and a reference impedance spectrum for the branch is constructed; during operation, temperature compensation is performed based on impedance changes to obtain dynamic branch temperature distribution data; further, the duty cycle of each branch is corrected based on the electro-thermal coupling relationship to form zoning modulation data; finally, matrix-based dynamic zoning PWM control is achieved through bistable voltage waveform analysis to suppress local hot spots and improve overall luminous uniformity.

[0005] For example, Chinese invention patent CN114143931B discloses an LED driving circuit, which includes a power switch, a sampling resistor, a control module, and a driving module. The control module generates a valid sampling control signal at the end of the first switching cycle and the beginning of the second switching cycle of the power switch, and generates a valid switching cycle control signal in the first two cycles of the power switch. The driving module is used to generate a driving signal to control the power switch to operate for at least two switching cycles, and to generate a valid comparison trigger signal when the sampling voltage is greater than the compensation voltage.

[0006] For example, Chinese invention patent application CN120730571A discloses a matrix-type dynamic partitioned PWM control device and method, which includes: analyzing the parallel branch control nodes of an LED array and calculating the initial impedance value of the branch to obtain the branch reference impedance spectrum; continuously driving the LED array during normal operation and compensating for temperature values ​​through the branch reference impedance spectrum to obtain dynamic branch temperature distribution data; performing electro-thermal duty cycle compensation for each branch in the LED array based on the dynamic branch temperature distribution data to generate partitioned modulation data; performing bistable voltage waveform analysis based on the partitioned modulation data, and then realizing matrix-type dynamic partitioned PWM control.

[0007] The above-mentioned technology has at least the following technical problems:

[0008] However, the aforementioned multi-source input data comes from different control cycles or different sampling mechanisms, resulting in inherent inconsistencies in their time scale. When they are sent to the controller, if they fail to be precisely aligned within the same control cycle, the calculated dimming boundary will deviate. This deviation is directly transmitted to the execution quantity, which may cause brightness jumps in adjacent zones or some zones to be too dark, or even cause overcurrent in the branch and protection actions, ultimately making it impossible to stably achieve the preset energy-saving target. Summary of the Invention

[0009] In view of this, embodiments of the present invention provide a method and system for energy-saving control of landscape LED light arrays based on partitioned load balancing, the specific technical solution of which is as follows:

[0010] Firstly, a landscape LED light array energy-saving control method based on partitioned load balancing is provided. The method includes: dividing the LED light array into multiple independent partitions, and pre-configuring a set of adjacent partition relationships, a branch current upper limit, and a brightness maintenance lower limit for each partition; within each control cycle, predicting the power of the next cycle based on historical data, and obtaining the target demand and real-time branch current margin from the landscape program instructions; and synchronizing the power of the next cycle, the target demand, and the real-time branch current margin with the start time of the current cycle as a reference to form an aligned data triplet.

[0011] Based on the aligned data triples, generate the upper and lower brightness boundaries for each partition, and perform boundary renegotiation when the lower boundary is greater than the upper boundary.

[0012] The instantaneous total power peak is analyzed based on the upper boundary of all partitions. If the peak exceeds the safety threshold, the timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the typical target brightness of each partition is directly determined as the final brightness of the corresponding partition, and the final brightness of each time period is converted into a drive signal output.

[0013] After the drive signal is executed, the actual current, brightness and gradient deviation are collected, and the boundary parameters and prediction model for the next cycle are corrected accordingly to form a closed-loop adaptive control.

[0014] Secondly, a landscape LED light array energy-saving control system based on partitioned load balancing is provided. This system includes: a partitioned configuration and data alignment module, a boundary generation and negotiation module, a peak control and timing scheduling output module, and a feedback acquisition and adaptive correction module.

[0015] The partition configuration and data alignment module is used to divide the LED light array into multiple independent partitions and pre-configure the adjacent partition relationship set, branch current upper limit and brightness maintenance lower limit for each partition. In each control cycle, the power of the next cycle is predicted based on historical data, and the target demand and real-time branch current margin from the landscape program instruction are obtained. The power of the next cycle, the target demand and the real-time branch current margin are synchronously aligned with the start time of the current cycle to form an aligned data triplet.

[0016] The boundary generation and negotiation module is used to generate the upper and lower brightness boundaries of each partition based on the aligned data triples, and to renegotiate the boundaries when the lower boundary is greater than the upper boundary.

[0017] The peak control and timing scheduling output module is used to analyze the instantaneous total power peak based on the upper boundary of all partitions. If the peak exceeds the safety threshold, the timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the typical target brightness of each partition is directly determined as the final brightness of the corresponding partition, and the final brightness of each time period is converted into a drive signal output.

[0018] The feedback acquisition and adaptive correction module is used to acquire the actual current, brightness and gradient deviation after the drive signal is executed, and correct the boundary parameters and prediction model for the next cycle accordingly to form closed-loop adaptive control.

[0019] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0020] 1. The landscape LED light array energy-saving control method based on partitioned load balancing provided by this invention precisely aligns the target demand, the power predicted for the next cycle based on historical data, and the real-time branch current margin with the start time of the current control cycle as the reference. This eliminates the time misalignment caused by different sampling periods and communication delays among the three types of data, ensuring that the calculation of the upper and lower boundaries of brightness is strictly based on complete information at the same moment. This avoids brightness jumps between adjacent partitions or some partitions being too dark due to data asynchrony, fundamentally solving the industry pain point of visual abruptness at best and triggering branch overcurrent protection at worst. Furthermore, the starting timing offset is allocated according to the size of the real-time current margin of each partition, so that the brightness of each partition changes smoothly from the lower boundary to the upper boundary within the cycle. Furthermore, the average brightness remains constant. Combined with time-slice discretization driving and overcurrent verification, the power peaks originally concentrated at the beginning of the cycle are evenly distributed throughout the entire control cycle. This achieves the suppression of power overload without affecting the visual effect of the landscape, and stably reaches the preset energy-saving index. In addition, by collecting actual current to convert equivalent brightness, the current deviation, brightness deviation and adjacent gradient deviation are calculated. The boundary parameters and power prediction of the next cycle are corrected according to the hierarchical coordination rule of prioritizing current safety, followed by gradient smoothness, and finally brightness accuracy. This forms a closed-loop adaptive control, which solves the long-term drift problem caused by dynamic load changes and model aging in the existing technology. It ensures that the system can still operate stably under complex working conditions and continuously meet the requirements of visual smoothness and electrical safety.

[0021] 2. This invention solves the problem of control reference disorder caused by sampling cycle conflict and communication jitter in traditional schemes by uniformly binding the target demand, predicted power and branch current margin to the start time of the current control cycle through differentiated strategies of nearest neighbor matching, periodic index correspondence and last sampling within the window, respectively. This improves the anti-interference capability and determinism of dimming boundary calculation.

[0022] 3. This invention collects the actual branch current at the end of the previous cycle, converts it into equivalent actual brightness through a drive calibration table, and uses this historical brightness as the key input for calculating the lower boundary of the current cycle. This transforms the spatial gradient constraint into a recursive relationship in the time dimension, so that the lower boundary of each partition is not only limited by its own minimum requirements, but also constrained by the actual state of the adjacent partitions in the previous cycle. This suppresses the gradient accumulation error caused by independent calculation in a single cycle, enhances the visual consistency between adjacent areas under long-term operation, and avoids sudden brightness changes.

[0023] 4. This invention adjusts local conflicts by renegotiation at the boundary, propagating them outward layer by layer through adjacency relationships. It uses current margin as the basis for adjusting weights, ensuring the fairness of the adjustment and preventing infinite loops. This solves the convergence problem when multiple constraints conflict simultaneously in complex network topologies, avoids the chain reaction of excessive dimming or local flickering that may be caused by traditional pruning methods, and ensures that a feasible boundary solution can be obtained under any topology, thereby maintaining the stable output of the entire light array. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart of a landscape LED light array energy-saving control method based on partitioned load balancing provided in this application embodiment;

[0026] Figure 2 This is a schematic diagram of the spatial distribution and partitioning of the LED light array involved in this embodiment;

[0027] Figure 3 This is a schematic diagram of the preset logic of the driver calibration table involved in this embodiment;

[0028] Figure 4 A comparison of actual power and predicted power, and a residual distribution diagram;

[0029] Figure 5 This embodiment includes a correlation diagram of the input features for partitioned power prediction and a calibration diagram of the upper boundary of brightness.

[0030] Figure 6 This is a schematic diagram of the energy-saving control system for landscape LED light arrays based on partitioned load balancing, provided in an embodiment of this application. Detailed Implementation

[0031] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0032] Embodiment 1 of the present invention Figure 1The flowchart of the landscape LED light array energy-saving control method based on partitioned load balancing provided includes: dividing the LED light array into multiple partitions; in each control cycle, acquiring the target demand, real-time branch current margin, and the power predicted for the next cycle based on historical data; aligning the three types of data to form a data triplet based on the start time of the current cycle; generating the upper and lower boundaries of brightness for each partition based on the aligned data; if the lower boundary is greater than the upper boundary, recursive boundary renegotiation is performed; estimating the instantaneous total power peak based on the upper boundaries of all partitions; if the peak exceeds the safety threshold, allocating the starting timing offset according to the current margin of each partition, so that the brightness of each partition changes smoothly from the lower boundary to the upper boundary within the cycle while maintaining the average brightness; otherwise, directly outputting the typical target brightness; mapping the target brightness in each time slice to the drive current through the drive calibration table and verifying overcurrent; synchronously outputting drive signals in the order of time slices; after execution, acquiring the actual current and converting it to equivalent brightness; calculating the current, brightness, and gradient deviation; correcting the boundary parameters and prediction model for the next cycle according to priority, forming a closed-loop adaptive control.

[0033] like Figure 2 This is a schematic diagram of the spatial distribution and partitioning of the LED light array involved in this embodiment, including: Figure 2 (A) is the original coordinate distribution map of the LED lights. The horizontal axis is longitude and the vertical axis is latitude. The scattered points represent the actual installation location of each LED light. Figure 2 (B) shows the zoning results. Adjacent lights that are spatially continuous and belong to the same power supply branch are grouped into independent zones on the same coordinate system. Each zone is marked with a different colored convex hull or closed curve to indicate its coverage area and is labeled with a zone number. The diagram shows that each zone covers several adjacent lights, and the zones are connected to each other through spatial adjacency. This diagram is used to verify the rationality of the zoning. The lights within each zone are spatially continuous and close to each other, and there are clear boundaries between different zones. This provides an intuitive spatial basis for subsequently configuring the set of adjacent zone relationships, the upper limit of branch current, and the lower limit of brightness maintenance for each zone. Meanwhile, Figure 2 The numbering order of the partitions in (C) is also used for subsequent data storage.

[0034] Each independent power supply circuit driving a spatially continuous set of LED lights in the landscape light array is defined as a zone. Specifically, a zone is defined as a set of LED lights whose geometric distance between adjacent lights is less than a distance threshold calibrated offline by professional technicians. This divides the LED landscape light array into multiple independent control zones. Each zone is assigned a unique zone number for accurate identification in subsequent data storage, addressing, and algorithm iteration. Based on this, three sets of basic parameters are configured for each zone. The first set is the adjacent zone relationship set, which records the numbers of all other spatially adjacent zones. This is used to apply output intensity gradient constraints between adjacent zones to prevent visual brightness abrupt changes. The second set is the branch current upper limit, which is set based on the hardware safety capacity of the drive circuit and the maximum withstand current of the LED chips. This limit restricts the actual current consumed by each zone during operation. The first group is the minimum forward current value, which is provided in the LED datasheet. It is defined as the minimum current required to maintain a linear relationship between the output light intensity of the LED and the driving current. When the actual driving current is lower than the minimum forward current value, the LED will enter the non-linear operating region, resulting in a decrease in luminous efficacy, color temperature shift, or even visible brightness jumps or flickering. Therefore, the minimum forward current value is read from the LED datasheet in advance and converted into the corresponding brightness value through the driving calibration table. This serves as the minimum brightness limit for this zone, ensuring that the output brightness of each zone will not be lower than the brightness value corresponding to the minimum forward current value during any dynamic adjustment process. This maintains the basic display quality and prevents the LED from entering the non-linear or flickering region. The above division and configuration are all completed at once during the LED startup phase, forming the static reference for all subsequent real-time control steps.

[0035] like Figure 3 This is a schematic diagram of the preset logic of the driver calibration table involved in this embodiment, including Figure 3 (D) represents the calibration relationship between the PWM duty cycle and the actual drive current. The calibration points in the figure were obtained by laboratory stepped current testing, and the relationship between the PWM (Pulse Width Modulation) duty cycle and the branch drive current was approximately proportional to the normal operating range through linear fitting. Figure 3 (E) represents the calibration relationship between the PWM duty cycle and the output brightness, indicating that the output brightness and the PWM duty cycle have a linear correspondence under constant current drive conditions. Thus, the equivalent linear mapping relationship between the brightness and the branch drive current can be further derived. Based on this mapping relationship, the reverse solution from the target brightness to the drive current can be achieved by looking up the calibration table during the operation phase, or the forward calculation of the equivalent brightness can be achieved by collecting the actual current, providing a unified parameter conversion basis for subsequent partitioned brightness scheduling and power constraint control.

[0036] The database's driver calibration table is pre-set for the LED chip model and driver circuit used in each partition. In a laboratory environment, a series of discrete input current values ​​are obtained through offline testing. Several test points are evenly selected between the minimum stable operating current and the upper limit of the branch current. For the current value at each test point, the corresponding PWM duty cycle and output brightness are measured. Then, a linear relationship model between the branch drive current, PWM duty cycle, and LED chip output brightness is fitted using the least squares method: with the PWM duty cycle as the independent variable, linear regressions are performed on the current and brightness through the origin, and the least squares method is used to fit the current... The slope of the two straight lines representing PWM duty cycle and brightness is the optimal proportional coefficient that minimizes the sum of the squares of the distances from the measured point to the line. By combining the above two linear relationships, it can be deduced that there is also a linear relationship between brightness and current. Finally, this linear model is written into the drive calibration table in the form of a lookup table and interpolation algorithm. It is used in real-time control to calculate the required current based on the target brightness or to convert the equivalent brightness based on the collected current. The drive calibration table is written into the database once during initialization. In all subsequent control processes, the required drive current is obtained by looking up the table in reverse based on the target brightness, or the equivalent brightness is obtained by looking up the table in forward based on the actual collected current.

[0037] At the start of each control cycle, the actual power sequence recorded in this partition during this cycle and the predicted power of this cycle stored at the end of the previous cycle are read from the storage unit. The sequence is then weighted and recursively calculated using an exponential smoothing algorithm to calculate the predicted power for the next control cycle. For the first control cycle after startup, since there is no historical power data available for smoothing, the average power measured in this partition within a preset time window is used as the initial state value of the predicted power. This prediction result will be used for subsequent power constraint judgment and boundary calculation.

[0038] The calculation formula for the exponential smoothing algorithm is: the predicted power for the next cycle is equal to the smoothing coefficient multiplied by the actual power of the current cycle, plus the difference between the smoothing coefficient and the predicted power of the current cycle, thereby achieving continuous prediction of the power change trend for the next control cycle.

[0039] The smoothing coefficient is determined by performing a step load test on the lamp array power supply system, measuring the time it takes for the actual power to change from its initial value to 63.2% of its final stable value, i.e., the time taken by the first-order inertial element. This time is recorded as the response time constant, which characterizes the dynamic response speed of the system and reflects how quickly it responds to power changes from the initial steady state to the final steady state. It belongs to the inherent physical time scale of the system determined by the characteristics of the drive circuit, constant current source, and load characteristics. The control period is defined as a fixed sampling update interval, and the control period length is recorded as a known value. Then, based on the response characteristics of the first-order low-pass filter, a definite smoothing coefficient is obtained, where α represents the smoothing coefficient, T represents the control period length, and t represents the response time constant.

[0040] like Figure 4 The comparison between actual and predicted power, and the residual distribution diagram, include: Figure 4 In (F), the horizontal axis represents the control period number, and the vertical axis represents the power value. The blue line represents the actual power, and the red line represents the predicted power for the next period given by the exponential smoothing method. Figure 4 (G) is a bar chart, with the horizontal axis representing the period number and the vertical axis representing the residual (actual power minus predicted power, unit: kW), used to quantify the prediction error for each period. The chart shows that the predicted power and actual power exhibit largely consistent trends, and the residuals are concentrated near zero without showing systematic positive or negative bias, indicating that the currently used exponential smoothing model is unbiased and meets the required accuracy.

[0041] The width of the preset time window is determined by consulting the device datasheet to obtain the sampling interval of the branch current sensor and the maximum communication delay from the sensor to the controller. Then, a professional technician sets the width of the time window to a multiple of the sum of the sampling interval and the communication delay. The specific multiple is determined by the professional technician, and the calculation result is rounded to the millisecond level. Then, with the start time of the current control cycle as the center, the window width is extended forward and backward by half to form a symmetrical time window. This time window is calculated once during initialization and stored in the database, and is applicable to all subsequent control cycles.

[0042] Within each control cycle, the target demand allocated to each zone for the current cycle is read from the program instructions. At the same time, the current actual branch current readback value of each zone is collected in real time by current sensors deployed on the branches. Then, based on the pre-configured upper limit of branch current, which is the maximum safe current allowed to pass through the drive circuit of that zone, the branch current margin is calculated, which is the upper limit of branch current minus the current actual branch current readback value. This margin represents the available capacity of the zone to increase current at the current moment.

[0043] If no valid demand data for a certain partition is received within the current control cycle, for example, due to lost program instruction transmission, communication interference, or parsing errors causing the target demand for that partition to be missing, the pre-configured lower limit of brightness maintenance for that partition is automatically read. The value of this lower limit of brightness maintenance is used as the conservative demand for that partition in subsequent boundary calculations and timing control. This ensures that even if a partition cannot receive explicit instructions from the upper layer temporarily, it can continue to operate based on the minimum brightness constraint, avoiding sudden changes in visual effects caused by a partition being completely turned off or excessive brightness gradients between adjacent partitions, while also buying time for instruction recovery.

[0044] If the actual branch current readback value of a certain zone cannot be obtained from the branch current sensor within the current control cycle, the real-time branch current margin of that zone cannot be calculated. In order to avoid the risk of overcurrent caused by blindly increasing the drive current when the current state is unknown, the branch current margin of that zone is automatically marked as non-amplifiable. During the period when the mark is effective, the output strength of that zone cannot be adjusted upward in all subsequent control operations. It can only maintain the existing output strength or be adjusted downward. Thus, at the cost of sacrificing the dynamic response capability of the zone, electrical safety is prioritized. After the branch current readback value is successfully obtained in the next cycle, the mark is automatically cleared and the normal amplification capability of the zone is restored.

[0045] The system receives the target demand for each partition, the predicted power for the next control cycle, and the real-time branch current margin. These three types of data are each accompanied by their own timestamps or cycle indices, indicating the time when they were generated or sampled. Due to communication delays, different sampling cycles, or asynchronous data processing, the times represented by these three types of data are not synchronized. In order to achieve unified decision-making within the control cycle, the start time of the current control cycle is used as the reference time point to align all data in time. Specifically, for the target demand, the nearest neighbor matching method is used: among all the received demand data with timestamps, the demand value whose timestamp is closest to the start time of the current cycle is selected as the effective demand for that partition in this cycle. In this way, the triplet data used for boundary calculation and timing control are consistent in time, avoiding control deviations caused by data misalignment.

[0046] The predicted power for the next cycle is assigned a corresponding cycle index during the calculation and generation process. Since there is a clear and fixed offset relationship between the index of the predicted power and the index of the current cycle, there is no need to perform timestamp matching or interpolation. The pre-calculated predicted power value can be read directly according to the index of the current cycle to complete the data alignment. This alignment method utilizes the natural ordinal relationship of the cycle index, avoids complex time matching calculations, and improves the determinism and efficiency of data synchronization.

[0047] The sampling time for branch current margin is not strictly synchronized with the start time of the control cycle. In order to ensure data timeliness while avoiding instability caused by instantaneous noise or sampling jitter, a preset time window width is set based on the start time of the current control cycle. Within this time window, all collected branch current margin sampling values ​​are retrieved, and the value with the latest timestamp, i.e., the last sampled value, is taken as the branch current margin used for boundary calculation and timing control in this cycle. If there are no sampling values ​​within the window, the most recent valid sampling value is used. This method effectively balances data freshness and sampling jitter robustness.

[0048] After the aforementioned time synchronization processing, the three types of data that might have arrived asynchronously or had different sampling times are ultimately bound to the same control cycle index, forming a complete and time-consistent data triplet. This triplet contains the target demand available in the current cycle, the power prediction for the next cycle, and the branch current margin for the current cycle, serving as the sole input basis for subsequent calculations of the upper and lower boundaries of brightness for each partition.

[0049] Figure 5 The correlation of input features and the upper boundary calibration map of brightness for partitioned power prediction involved in this embodiment include: Figure 5 (H) is the correlation matrix between each variable and the predicted power, used to guide the selection of input features for the exponential smoothing method: when the correlation coefficient between historical power and predicted power is high, this variable should be used as the input first; when the correlation between real-time branch current margin or target demand and predicted power is weak, its weight in the prediction model can be appropriately reduced, thereby simplifying the calculation and improving the prediction accuracy. Figure 5 (I) is a scatter plot of the real-time branch current margin and the upper boundary of brightness. By linearly fitting the measured upper boundary data of brightness of each partition under different current margins, the proportional coefficient of current margin to brightness upper limit is obtained to ensure the fairness and reproducibility of boundary adjustment.

[0050] It receives the aligned data triplet formed after time synchronization, and reads the pre-configured lower limit of brightness maintenance, upper limit of branch current, and preset drive calibration table for each partition from the storage unit. The calibration table records the correspondence between current value and pulse width modulation duty cycle, and the duty cycle is proportional to the output brightness.

[0051] The actual branch current value of each zone is collected by the branch current sensor. Since the drive calibration table records the PWM duty cycle corresponding to different current values ​​in advance, the corresponding duty cycle is obtained by looking up the table according to the collected actual current value. Since the duty cycle and the output brightness of the LED are directly proportional to the normal operating range, the equivalent actual brightness of each zone can be calculated by multiplying the found duty cycle by the preset proportional coefficient. The equivalent actual brightness is not directly measured, but indirectly calculated by the measurable current and the known mapping relationship. It is used for subsequent brightness lower boundary constraints, adjacent gradient deviation calculation and feedback correction.

[0052] The determination of the lower boundary of brightness needs to meet two basic constraints: First, the lower boundary must not be lower than the pre-configured lower limit of brightness maintenance to ensure the minimum visual effect; Second, the lower boundary also needs to consider the equivalent actual brightness of the adjacent partition in the previous cycle, which is calculated by the actual current fed back in the previous cycle through the drive calibration table, so that the absolute value of the difference between the lower boundary of the current partition and the equivalent actual brightness of the adjacent partition in the previous cycle does not exceed the preset maximum allowable brightness difference between adjacent partitions, thereby avoiding excessive brightness jumps between adjacent partitions. The larger of the lower limit of brightness maintenance and the maximum possible brightness allowed by the constraints of adjacent partitions is used as the lower boundary of brightness of the current partition. If there are multiple adjacent partitions, the one with the strictest constraint (i.e., the maximum value) among all adjacent partition constraints is compared with the lower limit of brightness maintenance, and the larger value of the two is taken as the lower boundary of brightness. If a partition has no adjacent partitions, the lower limit of brightness maintenance is used directly as its lower boundary of brightness. In this way, both the minimum visual output is ensured and abrupt brightness jumps between adjacent areas are avoided.

[0053] The preset logic for the maximum allowable brightness difference between adjacent zones is set by professional technicians as a ratio of the maximum output brightness of the light array. The size of this ratio is determined by professional technicians. This preset value is written into the configuration parameters during initialization and applies to all adjacent zones. Through the above method, those skilled in the art can uniquely adjust this parameter according to the actual application scenario, thereby ensuring a smooth and natural brightness transition between adjacent zones.

[0054] Based on the pre-configured upper limit of the branch current for this partition and referring to the preset drive calibration table, the maximum allowable brightness that this partition can achieve under the hardware limit of the branch current is calculated. Next, based on the predicted value of the total power for the next control cycle and the power budget already allocated to each partition, the maximum allowable brightness that this partition can bear under the total power constraint is analyzed, that is, the maximum brightness that this partition can be allocated without exceeding the total power. Finally, the two maximum allowable brightness values ​​and the target demand are compared, and the smaller one is taken as the upper limit of the brightness of this partition. The upper limit of the target demand brightness is used to ensure that the actual output brightness does not exceed the program control requirements. The upper limit of the branch current represents the physical safety boundary of the drive circuit, while the predicted total power and the allocated power budget represent energy scheduling constraints. All three are constraints that must be satisfied simultaneously. Therefore, the upper limit of brightness is determined by taking the minimum value to ensure that the program brightness requirements are met while taking into account the system power balance and drive electrical safety. This ensures that no constraint is violated under any operating conditions, thereby achieving system-level power balance scheduling while ensuring electrical safety.

[0055] When the calculated lower boundary of the brightness of a certain partition is greater than its upper boundary, it indicates that a feasible brightness range cannot be found for that partition under the current constraints. At this time, the boundary renegotiation mechanism is triggered. The specific steps of renegotiation are: the upper boundary of the brightness of the partition is forcibly adjusted to be equal to its lower boundary, that is, the brightness of the partition is fixed at the lower boundary value, and no margin is left.

[0056] Then, all partitions directly adjacent to the current partition are collected to form the set to be adjusted. This is because when a partition is forced to raise its upper boundary to the lower boundary due to its lower brightness boundary being greater than its upper boundary, the partition will always output a relatively high fixed brightness within this cycle. At this time, if its adjacent partitions still retain their original high brightness upper boundaries, the brightness difference between it and this fixed high-brightness partition is very likely to exceed the preset maximum allowable brightness difference between adjacent partitions during certain time slices when the brightness of the adjacent partitions increases. This will result in a visually obvious brightness jump or gradient over-limit. To avoid this situation, it is necessary to appropriately reduce the brightness upper boundary of the adjacent partitions to limit the highest brightness they can reach within this cycle, so that the brightness difference between the adjacent partitions and the fixed high-brightness partition at any time is always controlled within the allowable range. Therefore, reducing the upper boundary of the adjacent partitions is not to solve the boundary contradiction of the original partitions themselves, but to ensure that the brightness gradient constraint between adjacent partitions can still be met after the contradiction is resolved, thus ensuring the smoothness of the overall display effect.

[0057] For each adjacent partition in the set to be adjusted, the current actual current is subtracted from the upper limit of the branch current to obtain its current branch current margin, and these margins are accumulated to obtain the total current margin of all partitions to be adjusted. Then, according to the proportion of the current margin of each adjacent partition to the total margin, the upper limit of the brightness of that adjacent partition is reduced. The reduction amount is the difference between the current upper limit of the brightness of the adjacent partition and its own lower limit of brightness multiplied by this proportion, which compresses its adjustable range. It is ensured that the reduced upper limit of brightness is not less than the lower limit of brightness of the adjacent partition itself. Partitions with larger current margins mean that they still have a large margin from the upper limit of the branch current, so their brightness adjustment space is relatively more sufficient, and they are given priority in the boundary renegotiation process. Compressing the upper brightness boundary of this type of partition can release more local power resources without affecting its safety. However, partitions with small current margins are already close to the branch current limit, and their brightness adjustment capability is weak. Therefore, by allocating the compression amount according to the current margin ratio, over-compression of sensitive partitions can be avoided, thereby improving the stability of overall boundary coordination. After the upper boundary is reduced, the lower brightness boundary of the current partition is the greater of the following two values: the lower limit of brightness maintenance and the maximum value obtained by subtracting the maximum allowable brightness difference between adjacent partitions from the current executable target brightness of all adjacent partitions. The current executable target brightness of adjacent partitions is the brightness of the corresponding adjacent partition after adjustment during the current renegotiation process. The upper boundary is checked, and it is determined whether any new lower boundary of these adjacent partitions is greater than the upper boundary. If not, the renegotiation ends. If so, the adjacent partition is treated as a new conflicting partition, and the boundary renegotiation is recursively repeated. During the recursion, to avoid infinite loops, each partition is processed only once. The upper boundary of an already adjusted partition is not lowered again, but the lower boundary needs to be recalculated. If the recursion depth exceeds the maximum number of renegotiations calibrated offline by professional technicians, the recursion is terminated immediately, and the upper boundary of the brightness of all partitions in the entire area currently connected by adjacency relationships is uniformly and forcibly set to their respective current lower boundary. Then the renegotiation ends. This means that these partitions in this control cycle... During this period, the upward space for brightness adjustment will be lost, and only a fixed lower boundary value can be output. Although this approach will temporarily sacrifice some brightness dynamic range, it can ensure that convergence can still be fast even in the worst case. It avoids control cycle timeout due to excessive time consumption of boundary renegotiation. After the next control cycle begins, the boundary will be recalculated based on the new feedback data, and normal adjustment capability will be gradually restored. Through the above steps, boundary conflicts can be resolved within a limited number of iterations, ensuring that each partition has a feasible range where the lower boundary is no greater than the upper boundary. At the same time, partitions with smaller current margins are prioritized for protection because they account for a small proportion of the total margin, resulting in a small reduction ratio and a small actual reduction amount, thereby avoiding over-compression of sensitive partitions.

[0058] After calculating the upper and lower luminance boundaries for each partition, the data triplets containing the target demand, predicted power, and current margin, which were previously aligned, are received. In order to obtain a deterministic luminance control target while satisfying multiple constraints, the arithmetic mean of the lower and upper luminance boundaries for each partition is taken. This average value is used as the typical target luminance for that partition. The lower boundary ensures the minimum visual requirements and adjacent gradient constraints, while the upper boundary is limited by current and power safety. The average value of the two is usually located in the middle of the feasible range, which leaves a certain margin for upward and downward adjustment and avoids the problem of being too dark or too bright that may be caused by directly selecting the lower or upper boundary. This provides a balanced reference value for subsequent timing offset allocation and drive signal conversion. In the extreme case where the lower boundary is equal to the upper boundary, the typical target luminance is naturally equal to this common value.

[0059] To assess whether the instantaneous total power peak that may occur within the current control cycle exceeds the safety threshold, the upper boundary of the brightness of each partition is considered as the maximum output brightness that each partition can achieve under the premise of satisfying current, power, and adjacent gradient constraints. Since brightness is proportional to drive current, the drive current corresponding to the upper boundary of brightness for each partition can be obtained by reverse mapping through the drive calibration table. Therefore, the upper boundary of brightness of each partition is converted into the corresponding drive current, and combined with the known supply voltage, the instantaneous power of each partition under the worst case is calculated. Finally, the instantaneous power of all partitions is summed to obtain the peak instantaneous total power of the entire lamp array. Since the upper boundary of brightness represents the highest brightness that the partition is allowed to reach in this cycle, using the upper boundary of brightness for estimation and using this worst case as the evaluation basis can ensure that the actual total power will not exceed the peak estimate under any actual operating conditions, thus providing a conservative but safe judgment benchmark for subsequent timing offset control.

[0060] The calculated instantaneous total power peak is compared with a preset safety threshold. If the instantaneous total power peak is less than or equal to the safety threshold, it indicates that even if all partitions reach their respective brightness upper boundaries simultaneously, the total power will not exceed the limit. Therefore, there is no need for timing peak scheduling. At this time, the scheduling enable flag is marked as disabled, indicating that no timing offset needs to be allocated in this cycle. At the same time, the final brightness of each partition is directly determined to be equal to the previously calculated typical target brightness. That is, when the power margin is sufficient, the output is in the simplest offset-free mode, avoiding unnecessary timing calculations and drive delays, improving response efficiency and output stability. The scheduling enable flag is used to indicate whether partition power timing scheduling is started in the current control cycle. When the scheduling enable flag is enabled, partition power timing scheduling is performed on each partition. When the scheduling enable flag is disabled, each partition outputs directly according to the typical target brightness.

[0061] The safety threshold is determined by professional technicians who obtain the rated output power of the power supply for the light array and then preset a coefficient less than one based on safety redundancy requirements. This coefficient is determined by professional technicians through prior calibration experiments. Under the premise of ensuring that the power supply does not trigger overload protection, an empirical ratio is selected after comprehensively considering measurement errors, instantaneous fluctuations, and long-term reliability. The rated power is multiplied by this coefficient to obtain the safety threshold. In this way, a reasonable buffer margin is left for the power supply, and unnecessary timing scheduling is avoided due to the threshold being set too low, or the protection function is lost due to it being set too high. Through the above method, those skilled in the art can uniquely adjust the safety threshold according to the power supply parameters, light array load characteristics, and control cycle duration.

[0062] When the instantaneous total power peak exceeds the preset safety threshold, it indicates that if all zones simultaneously output according to their respective upper brightness boundaries, the total power will exceed the limit. Therefore, it is necessary to perform timing-based peak scheduling on the brightness output of each zone to reduce the instantaneous peak. At this time, the scheduling enable flag is set to enabled, the zone power timing scheduling module is started, and the three key parameters of each zone's lower brightness boundary, upper brightness boundary, and typical target brightness are passed to the zone power timing scheduling module as input for subsequent timing offset allocation and brightness trajectory planning. Different initial timing offsets will be allocated according to the branch current margin of each zone, so that the brightness of each zone changes smoothly from the lower boundary to the upper boundary within the cycle, while keeping the average brightness equal to the typical target brightness, thereby maintaining the consistency of visual effect as much as possible while meeting the safety threshold.

[0063] When the scheduling enable flag is set to enabled, the current control cycle is discretized in time, dividing the entire control cycle into multiple equal-length micro-time slices. The length of each time slice is obtained by dividing the control cycle duration by the preset total number of time slices. The purpose of this division is to support subsequent timing offset allocation, so that each partition can use the time slice as the smallest scheduling unit and start its brightness from the lower boundary to the upper boundary in different starting time slices. This disperses the total power peak that was originally concentrated at the beginning of the cycle to various time slices, reducing the instantaneous total power peak. The shorter the time slice length, the finer the scheduling granularity, and the better the peak shaving effect.

[0064] The total number of time slices is determined offline by professional technicians. The specific operation involves building an offline simulation platform that is exactly the same as the actual control. For a set of candidate total time slices, the limit test sequence of all partitions from the lower boundary to the upper boundary is run. The instantaneous total power peak, the number of adjacent brightness gradient exceedances, and the controller's computational load are recorded for each candidate value. Candidate values ​​that meet the requirements of instantaneous power not exceeding the safety threshold and zero gradient exceedances are selected. Then, the candidate value with the minimum computational load is selected as the optimal total number of time slices. This determined value is written into the database. Through this offline calibration, the optimal total number of time slices can be obtained for specific hardware and load characteristics, which ensures power safety and visual smoothness while avoiding the waste of computational resources.

[0065] All partitions are sorted from minimum to maximum based on their real-time branch current margins. Then, a starting timing offset is assigned to each partition: the partition with the smallest sequence number receives the preset minimum starting timing offset, the partition with the second smallest sequence number receives the offset plus one unit step, and so on, with the partition with the largest sequence number receiving the preset maximum starting timing offset. The starting timing offset represents the time elapsed from the start of the cycle to when the partition begins to rise from the lower boundary. The unit step is the control cycle duration divided by the total number of partitions. In this way, the starting timing offsets of each partition increase sequentially from smallest to largest current margin, and the offset sequence forms an arithmetic progression. This ensures that partitions with smaller current margins begin their brightness change process earlier, thereby distributing the power peak throughout the entire cycle. This allocation method does not require complex weighted calculations, but only relies on the sorting results and simple arithmetic operations, making it easy to implement in engineering and highly deterministic.

[0066] Each partition maintains its brightness at the lower boundary until its initial timing offset is reached. When the corresponding offset is reached, a linear ramp brightness change is performed with the lower brightness boundary as the starting value and the upper brightness boundary as the ending value. The slope of the linear change is adaptively adjusted according to the remaining control cycle duration, so that the average brightness of the partition within the control cycle is equal to the arithmetic mean of the lower and upper brightness boundaries. The upper boundary is reached exactly at the end of the cycle. This linear change automatically ensures that the average brightness within the entire cycle is equal to the typical target brightness. Then, each time slice is processed sequentially according to the time slice order: For the current time slice, the instantaneous brightness value of each partition at the beginning of each time slice is pre-calculated according to the rule of linear change of brightness of each partition with time. These brightness values ​​are organized into a two-dimensional table with the time slice number as the row and the partition number as the column, which is called the timing scheduling table.

[0067] In this embodiment, the soft-start curve is implemented as a linear ramp by determining the start time offset and the end time of the cycle for each partition. The brightness of the partition at the start time is known as the lower brightness boundary, and the brightness at the end time of the cycle is known as the upper brightness boundary. Then, the slope of the brightness change over time is calculated, which is the difference between the upper and lower brightness boundaries divided by the difference between the cycle duration and the start offset time. At any time after the start offset time and before the end of the cycle, the current brightness is equal to the lower brightness boundary plus the slope multiplied by the difference between that time and the start offset time. Before the start offset time, the brightness is constant at the lower brightness boundary. In this way, the brightness of each partition rises linearly from the start offset time to the upper brightness boundary at the end of the cycle, and the average brightness during the entire cycle is exactly equal to the average of the lower and upper brightness boundaries.

[0068] After allocating the initial offset time for each partition according to the branch current margin from smallest to largest, a gradient check is performed on each pair of adjacent partitions. At the beginning of each time slice, the absolute value of the instantaneous brightness difference between any two adjacent partitions does not exceed the preset maximum allowable brightness difference between adjacent partitions. For any two adjacent partitions, assuming that the partition with the smaller current margin starts rising first and its initial offset time is earlier, while the partition with the larger current margin starts offsetting later, the instantaneous brightness of these two partitions is calculated at the beginning of each time slice. If there exists any time slice where the absolute value of the brightness difference between these two partitions exceeds the maximum allowable brightness difference between adjacent partitions, the gradient is considered to have exceeded the limit. To address this... The problem is to keep the initial offset time of the partition with small current margin unchanged, while increasing the initial offset time of the partition with large current margin by one time slice, thereby delaying the start of brightness rise. This is because the brightness change process of each partition is determined by the initial timing offset. When the initial timing offset changes, the brightness-time function of the corresponding partition shifts along the time axis, changing the instantaneous brightness value corresponding to each time slice within the same control cycle. Specifically, increasing the initial timing offset by one time slice is equivalent to shifting the entire brightness rise process of that partition backward by one time slice, thus reducing the instantaneous brightness of that partition within the same time slice. Conversely, decreasing the initial timing offset by one time slice is equivalent to... The brightness increase process is shifted forward as a whole, thereby increasing the instantaneous brightness of the partition within the same time slice. Since the gradient constraints of adjacent partitions are determined based on the difference in instantaneous brightness within the same time slice, adjusting the initial timing offset can change the brightness difference between adjacent partitions within the same time slice, thus eliminating gradient overshoot and satisfying constraints. After adjustment, the brightness difference in all time slices is recalculated. If there is still an overshoot, the initial offset time of the partition with a larger current margin is increased by another time slice until the gradient meets the requirements. If the initial offset time of the partition with a larger current margin has been postponed to the end of the control cycle, meaning that adding another time slice to its initial offset time would exceed the cycle length, but still... If the gradient constraint cannot be met, the partition will no longer be postponed. Instead, the starting offset time of the partition with smaller current margin will be reduced by one time slice length each time, i.e., advanced by one time slice. However, it must be ensured that the reduced starting offset time is not less than zero, and the order of the starting offset times of all partitions after adjustment should still be consistent with the order of current margin from smallest to largest. That is, the offset time of the partition with smaller margin cannot be greater than that of the partition with larger margin. If the offset times of the two are equal or the gradient cannot be met even after reaching the boundary, the current gradient exceeds the limit or the maximum allowable brightness difference between adjacent partitions is reduced and recalculated. Through this iterative adjustment with a fixed step size of one time slice, the visual smoothness between adjacent partitions at any time is guaranteed while satisfying the power peak constraint.

[0069] After completing the timing offset allocation and gradient verification, a timing schedule table is generated. This table records the target brightness value of each partition in each time slice, which is the instantaneous brightness calculated based on the soft-start curve. When this timing schedule table is received, each time slice is processed one by one: for the current time slice, the target brightness value of each partition in the time slice is extracted, and then the inverse mapping function of the driver calibration table is used. Inverse mapping is to deduce the estimated drive current required to generate the brightness based on the known target brightness value, and convert the upper-level brightness control target into executable current drive parameters to ensure that the drive signal is consistent with the brightness requirement.

[0070] For each partition within each time slice, the target brightness value of the partition is back-mapped to an estimated drive current using a drive calibration table. This estimated drive current is then compared with the pre-configured upper limit of the branch current for that partition. If the estimated drive current does not exceed the upper limit of the branch current, the instantaneous brightness value for that time slice remains unchanged, and the process proceeds directly to the subsequent output stage. If the estimated drive current exceeds the upper limit of the branch current, the instantaneous brightness value for that time slice is reduced by the following ratio: the original target brightness value is multiplied by the ratio of the upper limit of the branch current to the estimated drive current to obtain the reduced brightness value. This reduced brightness value is then back-mapped to a new estimated drive current using the drive calibration table and compared again with the upper limit of the branch current. If the new estimated drive current still exceeds the upper limit, the above proportional reduction operation is repeated until the estimated drive current is less than or equal to the upper limit of the branch current. The corresponding brightness value at this point is the final instantaneous brightness value adopted by that partition within that time slice. This iterative verification mechanism ensures that the actual drive current of each partition in any time slice will not exceed the upper limit of the branch current allowed by the hardware, thereby ensuring electrical safety while achieving brightness control.

[0071] If the estimated drive current does not exceed the branch current limit, it indicates that the target brightness within the time slice will not cause an overcurrent risk. Therefore, the instantaneous brightness value of the time slice remains unchanged without any proportional reduction processing. The brightness value is directly used as the basis for calculating the final drive signal, ensuring that the original brightness plan is maintained first without exceeding the limit, avoiding unnecessary brightness reduction, and thus preserving the expected visual effect of the landscape program to the greatest extent.

[0072] After the estimated drive current is verified and the brightness is corrected for all time slices in sequence, the final safe instantaneous brightness value of each partition in each time slice is obtained. Then, the safe instantaneous brightness value is converted into the corresponding drive current or pulse width modulation duty cycle signal through the drive calibration table. Subsequently, the drive signals are output in ascending order of time slice number: at the beginning of each time slice, the drive signals corresponding to all partitions in the time slice are sent to the corresponding partition actuators at the same time, so that each partition maintains the brightness output determined by the drive signal in the time slice. After the end of a time slice, the start of the next time slice is automatically entered, and the above output process is repeated until all time slices in the current control cycle have been processed. This ensures that the instantaneous total power peak is effectively distributed across the time slices and that the brightness output of each partition in each time slice is synchronized in time and does not exceed the branch current limit.

[0073] When the scheduling enable flag is disabled, the final brightness of each partition is directly determined as the typical target brightness. Upon receiving this typical target brightness value, time slice division and timing offset allocation are no longer performed. Instead, the typical target brightness is directly converted into the corresponding PWM duty cycle through the inverse mapping process of the drive calibration table. Since the drive calibration table pre-records the correspondence between current and duty cycle, and the duty cycle is proportional to the brightness, the required duty cycle value can be uniquely determined based on the typical target brightness. This duty cycle value is converted into a constant drive signal and continuously output to the partition throughout the entire control cycle, enabling the partition to operate with a constant brightness.

[0074] After the drive signal output is completed, the actual branch current value of each partition is collected by the branch current sensor. Then, using a preset drive calibration table, the actual current is converted into the equivalent actual brightness of each partition. Next, based on the equivalent actual brightness of adjacent partitions, the absolute value of the difference between adjacent equivalent brightness is calculated. Based on this, three types of deviations are calculated: First, current deviation, which is obtained by subtracting the upper limit of the branch current of the partition from the actual branch current; second, brightness deviation, which is obtained by subtracting the typical target brightness of the partition from the equivalent actual brightness; and third, adjacent gradient deviation, which is obtained by subtracting the maximum allowable brightness difference between adjacent partitions from the absolute value of the difference between the equivalent actual brightness of adjacent partitions. These three types of deviations will be used for subsequent boundary parameter correction and prediction model update to form closed-loop adaptive control.

[0075] If the current deviation is greater than or equal to zero, meaning the actual current reaches or exceeds the upper limit, it indicates that the actual current is out of limit, meaning the drive circuit faces an overcurrent risk. Brightness increase must be immediately limited to protect the hardware. In the next control cycle, the upper limit of brightness for that zone is tightened by the amount equal to the preset current adjustment coefficient multiplied by the absolute value of the current deviation. Simultaneously, a "no brightening" flag is set for that zone, and its brightness cannot be adjusted upwards while the flag is active. If the current deviation is less than zero, meaning the actual current has not reached the upper limit, it indicates that the zone still has a safety margin. Appropriately widening the upper limit can restore its brightness adjustment capability and avoid limiting the display effect. In the next control cycle, the upper limit of brightness for that zone is widened by the same preset current adjustment coefficient multiplied by the absolute value of the current deviation. Through this bidirectional adjustment, brightness can be quickly limited in case of overcurrent, and brightness adjustment capability can be restored when there is a safety margin, thus balancing safety and dynamic response.

[0076] When the brightness deviation is greater than or equal to zero, it indicates that the current output brightness of the partition has reached or exceeded the typical target brightness, showing a tendency to be too high. At this time, in the next control cycle, a cooperative boundary correction operation is performed on the partition: based on the absolute value of the brightness deviation, the upper boundary of the brightness is adjusted downward by multiplying by the preset upper boundary adjustment coefficient and the lower boundary adjustment coefficient, respectively, so as to simultaneously narrow the adjustable brightness range of the partition. Among them, the upper boundary adjustment coefficient and the lower boundary adjustment coefficient are both preset positive parameters, and the upper boundary adjustment coefficient is greater than the lower boundary adjustment coefficient, so that the suppression amplitude of the upper boundary of brightness is greater than the rise amplitude of the lower boundary, thereby prioritizing the limitation of the maximum output brightness and preventing the brightness from continuing to rise. At the same time, by tightening the brightness range, the average brightness of the partition is gradually reduced, so that it converges towards the typical target brightness. The above method realizes the boundary adaptive adjustment and average brightness regression control based on the target deviation.

[0077] If the brightness deviation is less than zero, meaning the actual equivalent brightness is lower than the typical target brightness, it indicates that the brightness of the zone is insufficient and the average brightness needs to be increased. In this case, only the lower boundary is raised while the upper boundary remains unchanged. This allows the average brightness to be increased by raising the minimum brightness without sacrificing the maximum brightness potential, while retaining the space for upward adjustment of the upper boundary. This avoids limiting the subsequent dynamic response due to excessive tightening. In this case, only the lower boundary of the brightness of the zone is raised by a preset lower boundary adjustment coefficient multiplied by the absolute value of the brightness deviation, while the upper boundary remains unchanged. This two-way adjustment strategy can compress both the upper and lower boundaries to reduce the average brightness when the brightness is too high, and only raise the lower boundary to make up for the deficiency when the brightness is too low, while retaining the adjustment space of the upper boundary, thereby effectively maintaining the brightness within the desired range.

[0078] The upper and lower boundary adjustment coefficients are determined through offline calibration by professional technicians. An offline simulation platform identical to the actual situation is built, a set of candidate coefficient combinations is selected, and a brightness step response test is run for each combination. The adjustment time and overshoot of the actual brightness reaching the target value are recorded. The coefficient combination with the shortest adjustment time and the overshoot not exceeding the preset maximum overshoot is selected as the optimal value. Finally, the selected upper and lower boundary adjustment coefficients are written into the database and remain unchanged in all control cycles, allowing those skilled in the art to adjust these two coefficients through offline calibration according to the above steps.

[0079] If the adjacent gradient deviation is greater than or equal to zero, it indicates that the actual adjacent equivalent brightness difference has reached or exceeded the maximum allowable brightness difference between adjacent zones. Without correction, this will cause abrupt changes in the spatial brightness distribution, disrupting visual continuity and leading to localized overbrightness or contrast imbalance. Therefore, in the brightness allocation of the next cycle, the zone on the higher equivalent brightness side of the excess zone pair is identified. Then, the absolute value of the adjacent gradient deviation is determined. To share the overall brightness reduction requirement, the average of the current equivalent brightness of the zone minus the lower limit of brightness maintenance is calculated, taking into account the proportion of the excess zone pairs that are relatively high in the excess zone pairs. The proportion of the sum of the average values ​​of the partition sets on the high equivalent brightness side is used to calculate the brightness reduction that the partition should bear. The brightness reduction is calculated by multiplying the total brightness reduction requirement by the average value of the partition and dividing by the sum of the average values ​​of all adjustable partitions. Finally, the brightness reduction is subtracted from the instantaneous brightness value of the partition to obtain the reduced brightness value. The reduced brightness value is used as the new upper limit of brightness for the partition, which is the highest brightness that the partition is allowed to reach in subsequent cycles. If the adjacent gradient deviation is less than zero, that is, the gradient has not exceeded the limit, then the brightness of any partition is not forcibly adjusted, and it is allowed to gradually approach the upper limit of the gradient in subsequent cycles based on other constraints.

[0080] Based on the actual branch current of each zone and the known supply voltage, the actual total power of the current control cycle is calculated, which is the sum of the products of the actual current and voltage of each zone. Then, the power deviation is obtained by subtracting the power value of the next cycle predicted by the exponential smoothing method from this actual total power. This power deviation is fed back to the power prediction to update the smoothing coefficient in the exponential smoothing method online. Specifically, the power deviation is normalized according to the rated power to obtain the normalized power deviation. On this basis, the exponential smoothing coefficient is adaptively updated using the normalized power deviation. The new smoothing coefficient is equal to the old smoothing coefficient plus the normalized power deviation and a preset ratio. The product of factors, where the scaling factor is determined offline by professional technicians, adjusts the direction of increase or decrease of the smoothing coefficient according to the positive or negative direction of the power deviation: when the power deviation is positive, the smoothing coefficient is increased to enhance the response speed of the prediction model to the power increase trend; when the power deviation is negative, the smoothing coefficient is decreased to weaken the response sensitivity of the prediction model to the power decrease trend. The above update process is repeated in each control cycle, thereby realizing the online adaptive adjustment of the exponential smoothing prediction model, enabling it to continuously track the actual power change trend. The above update is performed once in each control cycle, thereby realizing online adaptive adjustment, allowing those skilled in the art to adjust according to this rule.

[0081] When multiple deviations, including current deviation, brightness deviation, and adjacent gradient deviation, are triggered simultaneously, and their respective required boundary adjustment directions conflict, coordination is carried out according to the following priority order: Current safety is the highest priority, that is, the tightening rule of current deviation is executed unconditionally first to ensure that the actual current never exceeds the upper limit of the branch current; then, within the boundary range allowed by current safety, the adjustment operations of brightness deviation and gradient deviation are performed. If the adjustment requirements of brightness deviation and gradient deviation conflict at this time, the constraint of gradient deviation is given priority, that is, the upper boundary of brightness of the over-limit side zone is reduced and the brightness reduction is shared according to the gradient deviation rule first, and then the brightness deviation is processed according to the remaining adjustment space; for brightness deviation that cannot be fully compensated due to priority restrictions, it is gradually compensated in subsequent control cycles. Through this hierarchical coordination mechanism, visual smoothness is maintained first under the premise of ensuring electrical safety, and finally the gradual convergence of brightness accuracy is achieved.

[0082] like Figure 6 The schematic diagram of the landscape LED light array energy-saving control system based on partition load balancing provided in the embodiments of this application includes: a partition configuration and data alignment module, a boundary generation and negotiation module, a peak control and timing scheduling output module, and a feedback acquisition and adaptive correction module.

[0083] The partition configuration and data alignment module is used to divide the LED light array into multiple independent partitions and pre-configure the adjacent partition relationship set, branch current upper limit and brightness maintenance lower limit for each partition. In each control cycle, the power of the next cycle is predicted based on historical data, and the target demand and real-time branch current margin from the landscape program instruction are obtained. The power of the next cycle, the target demand and the real-time branch current margin are synchronously aligned with the start time of the current cycle to form an aligned data triplet.

[0084] The boundary generation and negotiation module is used to generate the upper and lower brightness boundaries of each partition based on the aligned data triples, and to renegotiate the boundaries when the lower boundary is greater than the upper boundary.

[0085] The peak control and timing scheduling output module is used to analyze the instantaneous total power peak based on the upper boundary of all partitions. If the peak exceeds the safety threshold, the timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the typical target brightness of each partition is directly determined as the final brightness of the corresponding partition, and the final brightness of each time period is converted into a drive signal output.

[0086] The feedback acquisition and adaptive correction module is used to acquire the actual current, brightness and gradient deviation after the drive signal is executed, and correct the boundary parameters and prediction model for the next cycle accordingly to form closed-loop adaptive control.

[0087] In Embodiment 2 of the present invention, based on the unchanged aspects of Embodiment 1, all partitions can be sorted in ascending order of branch current margin: the initial timing offset can be allocated according to the current margin ratio. The specific analysis method is as follows:

[0088] Real-time current values ​​are read from the branch current sensors of each partition, and these real-time current values ​​are subtracted from the pre-configured upper limit of the branch current to obtain the real-time branch current margin of each partition. This margin represents the current capacity that the partition can still safely increase at the current moment. In order to allocate the time resources of the control cycle according to the current capacity of each partition, the current margins of all partitions are summed to obtain the total current margin. For each partition, its own current margin is divided by the total current margin to obtain a proportional coefficient between 0 and 1. This coefficient reflects the share of the partition in the total increaseable current capacity. Then, multiply the scaling factor by the total duration of the control cycle. The product is the starting timing offset of that partition. Since the sum of the scaling factors of all partitions is 1, the sum of the offsets of all partitions is exactly equal to the duration of the control cycle. Mapping the scaling factor to the duration of the control cycle naturally constitutes a proportional division of the entire cycle time axis. This makes the offset of each partition numerically equivalent to a complete division and redistribution of the control cycle, and each offset falls within the interval from zero to the duration of the cycle, requiring no additional normalization. Under this allocation rule, partitions with larger current margins receive larger offsets, meaning they are scheduled to start increasing brightness later in the cycle; partitions with smaller current margins receive smaller offsets, meaning they start increasing earlier. This is because partitions with smaller current margins are more prone to overcurrent due to power surges. Allowing them to start increasing and complete brightness changes earlier avoids overlapping with the power peaks of other partitions, thereby reducing the spike in the instantaneous total power. The peak power distribution is optimized so that the partitions with large current margins have sufficient capacity to withstand the later rise without causing overcurrent. Before their respective initial offset times, the brightness of each partition remains at the lower boundary of brightness. From the offset time, the brightness rises smoothly along a linear ramp, reaching the upper boundary of brightness at the end of the cycle. This linear change naturally ensures that the average brightness over the entire cycle is equal to the typical target brightness. This linear change process corresponds to a piecewise linear function, whose time average over the entire cycle is equal to the arithmetic mean of the upper and lower boundaries of the linear function. Therefore, the average brightness over the entire cycle is equal to the average of the lower and upper boundaries, i.e., the typical target brightness. Thus, without changing the peak power constraint, the balanced power distribution in the time domain is kept consistent with the average brightness. By allocating the offset according to the current margin value, the actual current capacity of each partition can be used more precisely to schedule the power peak, which protects the weak partitions and makes full use of the capacity of the rich partitions, thereby minimizing the power peak.

[0089] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A landscape LED light array energy-saving control method based on zoned load balancing, characterized in that, Includes the following steps: The LED light array is divided into multiple independent zones, and each zone is pre-configured with a set of adjacent zone relationships, a branch current upper limit, and a brightness maintenance lower limit. In each control cycle, the power of the next cycle is predicted based on historical data, and the target demand and real-time branch current margin from the landscape program instructions are obtained. The power of the next cycle, the target demand, and the real-time branch current margin are synchronized and aligned with the start time of the current cycle to form an aligned data triplet. Based on the aligned data triplet, generate the upper and lower brightness boundaries for each partition, and perform boundary renegotiation when the lower boundary is greater than the upper boundary; The instantaneous total power peak is analyzed based on the upper boundary of all partitions. If the peak exceeds the safety threshold, the timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the typical target brightness of each partition is directly determined as the final brightness of the corresponding partition, and the final brightness of each time period is converted into a drive signal output. After the drive signal is executed, the actual current, brightness and gradient deviation are collected, and the boundary parameters and prediction model for the next cycle are corrected accordingly to form a closed-loop adaptive control.

2. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The LED array is divided into multiple independent zones, and each zone is pre-configured with a set of adjacent zone relationships, a branch current upper limit, and a brightness maintenance lower limit. Within each control cycle, the power for the next cycle is predicted based on historical data. The specific analysis method is as follows: The LED landscape light array is divided into multiple zones according to its spatial location and power supply branches. Each zone is assigned a unique zone number. The set of adjacent zone relationships, the upper limit of branch current, and the lower limit of brightness maintenance are pre-configured for each zone. At the beginning of each control cycle, the historical power sequence of the past several cycles of this partition is read from the storage unit, and the predicted power of the next cycle is calculated using the exponential smoothing method. For the first control cycle, the average power is calculated within a preset time window and used as the initial state value of the exponential smoothing recursive formula.

3. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The specific analysis method for obtaining the target demand and real-time branch current margin from the landscape program instruction is as follows: The target demand for the current control cycle is read, and the current branch current readback value is read from the branch current sensor. Combined with the configured branch current upper limit, the branch current margin is analyzed. The target demand is used to characterize the target brightness value that each partition to be executed in the current control cycle is expected to achieve. If the target demand is missing, a conservative demand will be generated based on maintaining the lower limit of brightness. If the current branch current readback value is missing, the branch current margin of that partition is marked as non-amplifiable.

4. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The method involves synchronizing the power, target demand, and real-time branch current margin for the next cycle based on the start time of the current cycle, forming an aligned data triplet. The specific analysis method is as follows: Receive the target demand, the predicted power for the next cycle, the branch current margin, and the timestamp or cycle index attached to each of these three values. Based on the start time of the current control cycle, map the target demand to the instruction value whose timestamp is closest to the start time at that time through nearest neighbor matching. The predicted power for the next cycle itself has a cycle index, and it is directly aligned with the current cycle index. The branch current margin is taken as the value of the last sample within the preset time interval before and after the start time; After alignment, the three values ​​are bound to the same control cycle, forming an aligned data triplet.

5. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The process involves generating upper and lower brightness boundaries for each partition based on the aligned data triples, and renegotiating the boundaries when the lower boundary is greater than the upper boundary. The specific analysis method is as follows: Receive the aligned data triplet, and simultaneously read the pre-configured lower limit of brightness maintenance and upper limit of branch current, as well as the preset drive calibration table; The drive calibration table is used to record the pulse width modulation duty cycle corresponding to each partition under different input currents, and the duty cycle is directly proportional to the output brightness. Collect the actual branch current at the end of the previous cycle for each partition, find the duty cycle according to the drive calibration table, and then use the positive proportional relationship between the duty cycle and brightness to calculate the equivalent actual brightness of each partition. For each adjacent partition, the lower boundary of the current partition's brightness is the larger of the brightness preservation lower limit and the maximum value obtained by subtracting the maximum allowable brightness difference between adjacent partitions from the equivalent actual brightness of each of the adjacent partitions. For each adjacent partition, the upper limit of the brightness of the current partition is calculated based on the upper limit of the branch current of the partition and the drive calibration table to obtain the maximum allowable brightness under the current constraint. Then, based on the predicted total power of the next cycle and the power budget allocated to each partition, the maximum allowable brightness of the partition under the power constraint is analyzed. The smaller value between the two maximum allowable brightness and the target demand is taken as the upper limit of the brightness of the partition. If the calculated lower brightness boundary is greater than the upper brightness boundary, then perform boundary renegotiation: force the upper brightness boundary of the partition to be equal to its lower brightness boundary, and then process all partitions adjacent to the partition one by one. For each adjacent partition, reduce the upper brightness boundary of the adjacent partition according to the proportion of its current branch current margin to the total margin. The reduced upper brightness boundary shall not be less than the lower brightness boundary of the adjacent partition itself, and recalculate the lower brightness boundary of the adjacent partition. If, after adjustment, the lower boundary of the adjacent partition is still greater than the upper boundary, the above steps are recursively executed until the lower boundary of the brightness of all related partitions is no greater than its upper boundary. If the number of recursions exceeds the preset maximum number of renegotiations, then the upper brightness boundary of all partitions within the current connected region will be uniformly set to their respective lower brightness boundaries, and the renegotiation will end.

6. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The method involves analyzing the instantaneous total power peak based on the upper boundaries of all partitions. If the peak exceeds a safety threshold, a timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the brightness of each partition is directly adjusted from the lower boundary to the upper boundary without any timing offset. The specific analysis method is as follows: Receive the lower and upper boundaries of brightness, as well as the aligned data triplet, and take the average of the lower and upper boundaries of brightness as the typical target brightness. Take the upper boundary of brightness for all partitions as the worst-case estimate of the instantaneous total power peak; The power and brightness characteristics of each partition are collected, the upper boundary of the brightness of all partitions is substituted to obtain the power of each partition, and the sum is used to obtain the instantaneous total power peak. If the instantaneous total power peak is less than or equal to the preset safety threshold, the scheduling enable flag is set to disabled, and the final brightness is directly determined to be equal to the typical target brightness. If the instantaneous total power peak exceeds the preset safety threshold, the scheduling enable flag is set to enabled, and the lower brightness boundary, upper brightness boundary, and typical target brightness are passed to the partition power timing scheduler.

7. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 6, characterized in that: The scheduling enable flag is set to enabled, and the lower boundary of brightness, the upper boundary of brightness, and the typical target brightness are passed to the partition power timing scheduler. The specific analysis method is as follows: When the scheduling enable flag is enabled, the control cycle is divided into multiple small time slices of equal length. The length of each time slice is equal to the control cycle divided by the preset total number of time slices. All partitions are sorted in ascending order of branch current margin to obtain the sequence number of each partition in the sorting sequence. The partition with the smallest sequence number is assigned a preset minimum starting timing offset, the partition with the second smallest sequence number is assigned a starting timing offset that increases by one unit step, and so on, with the partition with the largest sequence number receiving a preset maximum starting timing offset. The unit step is equal to the control cycle duration divided by the total number of partitions. The instantaneous brightness value of each partition remains at the lower limit of brightness before its start time slice, and then increases linearly to the upper limit of brightness according to the preset soft-start curve starting from the start time slice, so that the average brightness of all partitions in the whole control cycle is equal to the typical target brightness. Meanwhile, if the initial timing offset difference between adjacent partitions is too small, causing the brightness difference within any time slice to exceed the pre-configured maximum allowable brightness difference between adjacent partitions, the offset of the larger margin partition will be increased until the gradient constraint is met.

8. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: The specific analysis method for converting the target brightness of each time period into a driving signal output is as follows: If a timing schedule table is received, for each time slice, the target brightness value of each partition within each time slice is reverse-mapped to the estimated drive current through the drive calibration table; When the estimated drive current exceeds the upper limit of the branch current of the partition, the instantaneous brightness value of the time slice is reduced by the ratio of the upper limit of the branch current to the estimated current, and the verification is repeated until the estimated current does not exceed the upper limit of the branch current. If the estimated drive current does not exceed the upper limit of the branch current of the partition, the instantaneous brightness value of the time slice remains unchanged. After performing the above verification and correction on all time slices in sequence, the drive signals corresponding to each partition within each time slice are simultaneously output to all partitions at the beginning of each time slice, according to the order of the time slices. If a typical target brightness is received, it is directly converted into a constant pulse width modulation duty cycle through the drive calibration table.

9. The energy-saving control method for landscape LED light arrays based on partitioned load balancing as described in claim 1, characterized in that: After executing the drive signal, the actual current, brightness, and gradient deviation are collected. Based on this, the boundary parameters and prediction model for the next cycle are corrected to form a closed-loop adaptive control. The specific analysis method is as follows: After executing the drive signal, the actual current of each partition is collected. The actual current is converted into the equivalent actual brightness of each partition through the preset drive calibration table in the database. The difference between adjacent equivalent brightness is calculated based on the equivalent actual brightness of adjacent partitions. In this way, the current deviation is equal to the actual branch current minus the upper limit of the branch current, the brightness deviation is equal to the equivalent actual brightness minus the lower limit of brightness maintenance, and the adjacent gradient deviation is equal to the absolute value of the difference between the actual brightness of adjacent partitions minus the maximum allowable brightness difference between adjacent partitions. If the current deviation is greater than or equal to zero, the upper boundary of the brightness of the partition will be tightened in the next cycle by the amount of tightening, which is the preset current coefficient multiplied by the absolute value of the current deviation, and a prohibition on brightening will be set. If the current deviation is less than zero, the upper boundary of the brightness will be relaxed in the next cycle by the amount of relaxation, which is the preset current coefficient multiplied by the absolute value of the current deviation. If the brightness deviation is greater than or equal to zero, then in the next cycle, the upper boundary of brightness will be lowered by a preset upper boundary adjustment coefficient multiplied by the absolute value of the brightness deviation, and the lower boundary of brightness will be raised by a preset lower boundary adjustment coefficient multiplied by the absolute value of the brightness deviation. Both adjustment coefficients are positive and preset by the system. If the brightness deviation is less than zero, then only the lower boundary of brightness will be raised by a preset lower boundary adjustment coefficient multiplied by the absolute value of the brightness deviation, while the upper boundary of brightness remains unchanged. If the adjacent gradient deviation is greater than or equal to zero, in the brightness allocation of the next cycle, identify the side of the high equivalent brightness of the over-limit partition, determine the gradient over-limit amount that needs to be eliminated, and then share the total brightness reduction demand according to the ratio of the average of the current branch current margin and the lower limit margin of the brightness of the partition to the average of all adjustable partitions. At the same time, the reduced instantaneous brightness value is used as the new upper limit of the brightness of the partition. If the adjacent gradient deviation is less than zero, the brightness of any partition is not forcibly adjusted. Simultaneously, based on the actual branch current of each partition and the known power supply voltage, the actual total power of the current control cycle is calculated, and the deviation between the actual total power and the predicted power is fed back to the power prediction, and the smoothing coefficient in the exponential smoothing method is updated online. When multiple deviations are triggered simultaneously and their adjustment directions conflict, current safety takes the highest priority. That is, the tightening rule of current deviation is executed first, and then the adjustment of brightness deviation and gradient deviation is executed within the allowable current range. If brightness deviation and gradient deviation conflict, the gradient deviation is satisfied first, and the brightness deviation is gradually compensated in subsequent cycles.

10. A landscape LED light array energy-saving control system based on zoned load balancing, characterized in that, include: Partition configuration and data alignment module, boundary generation and negotiation module, peak control and timing scheduling output module, and feedback acquisition and adaptive correction module; The partition configuration and data alignment module is used to divide the LED light array into multiple independent partitions, and pre-configure the adjacent partition relationship set, the upper limit of branch current and the lower limit of brightness maintenance for each partition. In each control cycle, the power of the next cycle is predicted based on historical data, and the target demand and real-time branch current margin from the landscape program instruction are obtained. The power of the next cycle, the target demand and the real-time branch current margin are synchronously aligned with the start time of the current cycle to form an aligned data triplet. The boundary generation and negotiation module is used to generate the upper and lower brightness boundaries of each partition based on the aligned data triplet, and to renegotiate the boundaries when the lower boundary is greater than the upper boundary. The peak control and timing scheduling output module is used to analyze the instantaneous total power peak based on the upper boundary of all partitions. If the peak exceeds the safety threshold, the timing offset is allocated according to the current margin of each partition. If the peak does not exceed the safety threshold, the typical target brightness of each partition is directly determined as the final brightness of the corresponding partition, and the final brightness of each time period is converted into a drive signal output. The feedback acquisition and adaptive correction module is used to acquire the actual current, brightness and gradient deviation after the drive signal is executed, and correct the boundary parameters and prediction model for the next cycle accordingly to form closed-loop adaptive control.

Images