Energy consumption optimization method applied to multi-screen integration

Abstract

The application discloses an energy consumption optimization method applied to multi-screen integration, relates to the technical field of screen integration, and comprises the following steps: a deep learning model for scoring the energy consumption of a single screen is constructed, and sample data are collected to train the deep learning model; a database for storing historical use data of the single screen is constructed, and the historical use data stored in the database is input into the trained deep learning model; according to the energy consumption evaluation score output by the deep learning model and the business demand of the integrated screen, corresponding single screens are selected, and the assembly sequence of the selected single screens is output; by using an LSTM network, temperature, voltage and other short-term non-structural interference are effectively filtered out through a forgetting gate, and features representing permanent loss such as accumulated time length and maintenance records are written into cell states through an input gate, so that surface fluctuations can be penetrated, and an evaluation score reflecting the real aging energy consumption level of the single screen can be accurately output, thereby solving the problem that the real power consumption of an old screen is difficult to quantify.

Description

Technical Field

[0001] This invention relates to the field of screen integration technology, specifically to a method for optimizing energy consumption in multi-screen integration. Background Technology

[0002] Multi-screen integration technology is widely used in commercial displays, automotive cockpits, and control and scheduling. Integrated screens are typically assembled from multiple discrete individual screens. In actual production and maintenance, companies often use a strategy of mixing new screens with recycled old screens to control costs. However, after long-term operation, the actual energy consumption of individual screens will experience an irreversible structural increase due to factors such as accumulated lighting time, extreme temperature shocks, and component aging.

[0003] Traditional integrated screen assembly methods rely solely on factory static parameters or simple size matching, failing to accurately assess the true power consumption of old or used individual screens throughout their current lifecycle. Blindly combining high-energy-consuming aging screens can easily lead to the overall power consumption of the integrated screen exceeding the system budget limit, causing problems such as localized overheating, power overload, and a sharp reduction in lifespan. Furthermore, existing power consumption assessment methods mostly focus on instantaneous power consumption monitoring, lacking long-term time-series mining of multi-dimensional historical operating data of the screens. This makes it difficult to eliminate short-term environmental interference to extract true permanent loss characteristics, resulting in poor energy efficiency consistency of the assembled integrated screens and difficulty in meeting complex and ever-changing customized business needs. Therefore, we now provide an energy consumption optimization method for multi-screen integration. Summary of the Invention

[0004] The purpose of this invention is to provide a power consumption optimization method for multi-screen integration.

[0005] The objective of this invention can be achieved through the following technical solution: an energy consumption optimization method applied to multi-screen integration, comprising: Step S1: Construct a deep learning model for scoring the energy consumption of a single screen, and collect sample data to train the deep learning model; Step S2: Construct a database to store historical usage data of individual screens, and input the historical usage data stored in the database into the trained deep learning model; Step S3: Based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen, select the corresponding single screen and output the assembly sequence of the selected single screen.

[0006] Furthermore, the deep learning model is a deep learning model constructed using an LSTM network, including: Cellular state, used as a baseline to characterize the cumulative aging features of a single screen; Forget gates are used to filter and remove short-term non-structural interference features in the cell state that are invalid for current energy consumption assessment. An input gate is used to identify and extract new features characterizing structural permanent loss from current historical data and write them into the cell state; The output gate is used to extract and output a comprehensive temporal feature vector reflecting the true energy consumption level of a single screen at the current time step, based on the updated cell state.

[0007] Furthermore, the process of training a deep learning model includes: Collect historical operating data corresponding to different types of single screens as sample data. The historical operating data includes actual power consumption values ​​under multi-dimensional features, including temperature, voltage, and brightness. The obtained sample data is divided into training set, test set and validation set; The training data is input into the LSTM network in chronological order, passing through the forget gate and the input gate to update the cell state, and finally the output gate outputs the predicted energy consumption evaluation score. The loss function uses mean squared error to obtain the difference between the predicted energy consumption assessment score and the actual value. The smaller the difference, the better the inspection result, and vice versa. Backpropagation is performed based on the obtained energy consumption assessment score to determine the weight parameters of each part; The weight parameters of each part are optimized using an optimizer; Repeat the above steps to iteratively train all sample data in the training set until the model converges, resulting in a fully trained deep learning model.

[0008] Furthermore, the process of constructing a database to store historical usage data for individual screens and inputting the historical usage data stored in the database into the trained deep learning model includes: Each standby screen is assigned a number, and a corresponding sub-database is built in the database based on each screen number, with each sub-database associated with a screen. Set up a data entry port to input the basic information and historical usage data of the standby single screen through the data entry port, and import the input historical usage data into the corresponding data sub-database. The historical usage data includes cumulative lighting time, highest / lowest temperature experienced, historical maintenance records, and drive parameters after each calibration. The basic information of the single screen includes its size and type; Historical usage data is input into the trained deep learning model, and the energy consumption evaluation score of each individual screen is output.

[0009] Furthermore, the business requirements include the target size range of the integrated screen, the display content characteristics, and the maximum total power consumption budget. The business requirements are prioritized as follows: target size range > maximum total power consumption budget > display content characteristics.

[0010] Furthermore, based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen, the process of selecting the corresponding single screen includes: Obtain the business requirements for integrated screens, and obtain the corresponding candidate single screen combinations based on the priority of the business requirements; Based on the energy consumption assessment scores of each individual screen in the candidate individual screen combination, the comprehensive evaluation coefficient of the candidate individual screen combination is obtained. The comprehensive evaluation coefficients of the various candidate single screen combinations are sorted from low to high, and the candidate single screen combination with the highest ranking is selected as the final single screen combination. The numbers of each single screen in the final single screen combination are output. Based on the output single screen numbers, the corresponding single screen is selected from the candidate single screens as the component of the integrated screen corresponding to the business requirement.

[0011] Furthermore, the process of obtaining the corresponding candidate single-screen combinations based on the priority of business needs is as follows: Based on the target size range in the business requirements, generate all first single-screen combinations that meet the business requirements; Based on the maximum total power consumption budget in the business requirements, the generated first single screen combination is filtered, and the single screen combination obtained by the filtering is recorded as the second single screen combination. Based on the display content characteristics in the business requirements, the obtained second single-screen combinations are filtered to obtain candidate single-screen combinations.

[0012] Beneficial effects 1. The energy consumption score output by deep learning is combined with the business requirements of integrated screens (size, total power consumption budget, content features), and a strict priority funnel screening mechanism is introduced. Under the premise of ensuring basic physical specifications and power consumption red lines, the single screen combination sequence with the lowest comprehensive evaluation coefficient (i.e. the best energy efficiency) is automatically calculated and selected, thus avoiding the risk of local overload after assembly from the source. 2. A customized LSTM network is adopted, which effectively filters out short-term unstructured interference such as temperature and voltage through the forget gate, and writes features representing permanent damage such as cumulative duration and maintenance records into the cell state through the input gate. This allows it to penetrate surface fluctuations and accurately output an evaluation score that reflects the true aging energy consumption level of a single screen, thus solving the problem of difficulty in quantifying the true power consumption of old screens. Attached Figure Description

[0013] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0014] The present invention will be further described below with reference to specific embodiments and accompanying drawings.

[0015] like Figure 1 As shown, the energy consumption optimization method applied to multi-screen integration includes: Step S1: Construct a deep learning model for scoring the energy consumption of a single screen, and collect sample data to train the deep learning model; Step S2: Construct a database to store historical usage data of individual screens, and input the historical usage data stored in the database into the trained deep learning model; Step S3: Based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen, select the corresponding single screen and output the assembly sequence of the selected single screen.

[0016] It should be further explained that, in the specific implementation process, the deep learning model is a deep learning model constructed using a Long Short-Term Memory (LSTM) network; The deep learning model includes: Cellular state, used as a baseline to characterize the cumulative aging features of a single screen; Forget gates are used to filter and remove short-term non-structural interference features in the cell state that are invalid for current energy consumption assessment. An input gate is used to identify and extract new features characterizing structural permanent loss from current historical data and write them into the cell state; The output gate is used to extract and output a comprehensive temporal feature vector reflecting the true energy consumption level of a single screen at the current time step, based on the updated cell state.

[0017] It should be further explained that the process of training a deep learning model includes: Historical operating data corresponding to different types of single screens are collected as sample data. The types of single screens include LCD screens and LED screens. The historical operating data includes actual power consumption values ​​under multi-dimensional features, such as temperature, voltage, and brightness. The obtained sample data is divided into training set, test set and validation set, usually in a ratio of 8:1:1; The training data is input into the LSTM network in chronological order, passing through the forget gate and the input gate to update the cell state, and finally the output gate outputs the predicted energy consumption evaluation score. The loss function uses mean squared error to obtain the difference between the predicted energy consumption assessment score and the actual value. The smaller the difference, the better the inspection result, and vice versa. Backpropagation is performed based on the obtained energy consumption assessment score to determine the weight parameters of each part; The optimizer optimizes the weight parameters of each part obtained by the optimizer. The optimizer can be represented as: W_new=W_old-γ• Loss / W; Where γ is the learning rate, W_new is the optimized weight parameter, W_old is the original weight parameter, and Loss is the loss function. Loss / W represents the partial derivative of the loss function with respect to the weight parameters; Repeat the above operations to iteratively train all sample data in the training set until the model converges, resulting in a trained deep learning model. A customized LSTM network is used, which effectively filters out short-term unstructured interference such as temperature and voltage through the forget gate, and writes features representing permanent damage, such as cumulative duration and maintenance records, into the cell state through the input gate. This allows the network to penetrate surface fluctuations and accurately output an evaluation score that reflects the true aging energy consumption level of a single screen, solving the problem of difficulty in quantifying the true power consumption of old screens.

[0018] It should be further explained that, in the specific implementation process, the process of constructing a database to store historical usage data of individual screens and inputting the historical usage data stored in the database into the trained deep learning model includes: Each standby screen is assigned a number, and a corresponding sub-database is built in the database based on each screen number, with each sub-database associated with a screen. Set up a data entry port to input the basic information and historical usage data of the standby single screen through the data entry port, and import the input historical usage data into the corresponding data sub-database. The historical usage data includes cumulative lighting time, highest / lowest temperature experienced, historical maintenance records, and drive parameters after each calibration. The basic information of the single screen includes its size and type; Historical usage data is input into the trained deep learning model, and the energy consumption evaluation score of each individual screen is output.

[0019] It should be further explained that, in the specific implementation process, the selection of the corresponding single screen based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen includes: Obtain the business requirements of the integrated screen, including the target size range of the integrated screen (width and height of the integrated screen), display content characteristics ("globally bright", "center bright and surrounding dark", "dynamic trajectory following") and maximum total power consumption budget; Prioritize the acquired business requirements, where target size range > maximum total power consumption budget > display content features; Based on the "target size range" in the business requirements, generate all single-screen combinations that meet the business requirements, and denot them as the first single-screen combination. Specifically, the total size after the combination is made to be within the target size range by superimposing the sizes of the single screens. Based on the maximum total power consumption budget in the business requirements, the generated first single screen combination is filtered, and the single screen combination obtained by the filtering is recorded as the second single screen combination. Specifically, the filtering is carried out by the sum of the actual power consumption values ​​of each single screen. Based on the display content characteristics in the business requirements, the obtained second single screen combination is filtered to obtain candidate single screen combinations. Specifically, the single screen type that can display the corresponding display content characteristics is selected through filtering. Each individual screen in the selected individual screen combination is labeled and denoted as i, where i = 1, 2, ..., n, n is an integer, and n > 1; The energy consumption assessment score corresponding to the single screen labeled i is denoted as . ; Based on the energy consumption assessment scores of each individual screen in each candidate single-screen combination, the comprehensive evaluation coefficient of the candidate single-screen combination is obtained, denoted as Kp, where: ; Among them, P P_max represents the average energy consumption assessment score of each individual screen in the candidate individual screen combination, P_min represents the maximum energy consumption assessment score in the candidate individual screen combination, and P_min represents the minimum energy consumption assessment score in the candidate individual screen combination. The comprehensive evaluation coefficients of the various candidate single-screen combinations are sorted from low to high, and the candidate single-screen combination with the highest ranking is selected as the final single-screen combination. The numbers of each single screen in the final single-screen combination are output. Based on the output single-screen numbers, the corresponding single screens are selected from the candidate single screens as the components of the integrated screen corresponding to the business requirement. The energy consumption score output by deep learning is combined with the business requirements of the integrated screen (size, total power consumption budget, content features), and a strict priority funnel screening mechanism is introduced. Under the premise of ensuring the basic physical specifications and power consumption red line, the single-screen combination sequence with the lowest comprehensive evaluation coefficient (i.e., the best energy efficiency) is automatically calculated and selected, thus avoiding the risk of local overload after assembly from the source.

[0020] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and various changes and modifications can be made without departing from the spirit and scope of the invention, all of which fall within the scope of the claims. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A power consumption optimization method applied to multi-screen integration, characterized in that, include: Step S1: Construct a deep learning model for scoring the energy consumption of a single screen, and collect sample data to train the deep learning model; Step S2: Construct a database to store historical usage data of individual screens, and input the historical usage data stored in the database into the trained deep learning model; Step S3: Based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen, select the corresponding single screen and output the assembly sequence of the selected single screen.

2. The energy consumption optimization method for multi-screen integration according to claim 1, characterized in that, The deep learning model is a deep learning model constructed using an LSTM network, including: Cellular state, used as a baseline to characterize the cumulative aging features of a single screen; Forget gates are used to filter and remove short-term non-structural interference features in the cell state that are invalid for current energy consumption assessment. An input gate is used to identify and extract new features characterizing structural permanent loss from current historical data and write them into the cell state; The output gate is used to extract and output a comprehensive temporal feature vector reflecting the true energy consumption level of a single screen at the current time step, based on the updated cell state.

3. The energy consumption optimization method for multi-screen integration according to claim 2, characterized in that, The process of training a deep learning model includes: Collect historical operating data corresponding to different types of single screens as sample data. The historical operating data includes actual power consumption values ​​under multi-dimensional features, including temperature, voltage, and brightness. The obtained sample data is divided into training set, test set and validation set; The training data is input into the LSTM network in chronological order, passing through the forget gate and the input gate to update the cell state, and finally the output gate outputs the predicted energy consumption evaluation score. The loss function uses mean squared error to obtain the difference between the predicted energy consumption assessment score and the actual value. The smaller the difference, the better the inspection result, and vice versa. Backpropagation is performed based on the obtained energy consumption assessment score to determine the weight parameters of each part; The weight parameters of each part are optimized using an optimizer; Repeat the above steps to iteratively train all sample data in the training set until the model converges, resulting in a fully trained deep learning model.

4. The energy consumption optimization method for multi-screen integration according to claim 3, characterized in that, The process of constructing a database to store historical usage data for individual screens and inputting the historical usage data stored in the database into a trained deep learning model includes: Each standby screen is assigned a number, and a corresponding sub-database is built in the database based on each screen number, with each sub-database associated with a screen. Set up a data entry port to input the basic information and historical usage data of the standby single screen through the data entry port, and import the input historical usage data into the corresponding data sub-database. The historical usage data includes cumulative lighting time, highest / lowest temperature experienced, historical maintenance records, and drive parameters after each calibration. The basic information of the single screen includes its size and type; Historical usage data is input into the trained deep learning model, and the energy consumption evaluation score of each individual screen is output.

5. The energy consumption optimization method for multi-screen integration according to claim 4, characterized in that, The business requirements include the target size range of the integrated screen, the characteristics of the display content, and the maximum total power consumption budget. The business requirements are prioritized as follows: target size range > maximum total power consumption budget > display content characteristics.

6. The energy consumption optimization method for multi-screen integration according to claim 5, characterized in that, The process of selecting the appropriate standalone screen based on the energy consumption assessment score output by the deep learning model and the business requirements of the integrated screen includes: Obtain the business requirements for integrated screens, and obtain the corresponding candidate single screen combinations based on the priority of the business requirements; Based on the energy consumption assessment scores of each individual screen in the candidate individual screen combination, the comprehensive evaluation coefficient of the candidate individual screen combination is obtained. The comprehensive evaluation coefficients of the various candidate single screen combinations are sorted from low to high, and the candidate single screen combination with the highest ranking is selected as the final single screen combination. The numbers of each single screen in the final single screen combination are output. Based on the output single screen numbers, the corresponding single screen is selected from the candidate single screens as the component of the integrated screen corresponding to the business requirement.

7. The energy consumption optimization method for multi-screen integration according to claim 6, characterized in that, The process of obtaining the corresponding candidate single-screen combinations based on the priority of business needs is as follows: Based on the target size range in the business requirements, generate all first single-screen combinations that meet the business requirements; Based on the maximum total power consumption budget in the business requirements, the generated first single screen combination is filtered, and the single screen combination obtained by the filtering is recorded as the second single screen combination. Based on the display content characteristics in the business requirements, the obtained second single-screen combinations are filtered to obtain candidate single-screen combinations.

Images