Multifunctional mobile power supply light engineering test system and method

Abstract

The invention discloses a multifunctional mobile power supply light engineering test system and method, and the method comprises the following steps: S1, initializing a test system, starting self-inspection, and confirming that equipment is in a to-be-tested state; s2, completing connection operation according to the lamp parameters and the protocol type; s3, activating a power supply module, selecting a power supply channel, and controlling output parameters; s4, starting a light control module, and setting a control protocol and signal parameters; s5, acquiring operation response data, and inputting the operation response data into the variational auto-encoder model; s6, optimizing the structure hyper-parameters of the variational auto-encoder model by using a beetle antennae search algorithm; s7, the reconstruction output is compared with the original data, abnormity is identified, and a report is generated; and S8, closing power supply and control signals, storing data and synchronizing the remote platform. According to the invention, intelligent power supply, response analysis and abnormity identification of lamp testing are realized through integration of an adjustable power supply, protocol identification control, data acquisition, variational auto-encoder modeling and longhorn beetle search optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and particularly to a multifunctional mobile power supply lighting project test system and method. Background Art

[0002] With the continuous improvement of the requirements for lamp control accuracy, response speed and operation stability in the fields of urban lighting projects, stage performances, building landscapes, tunnel lighting, etc., various high-performance lighting devices such as LED lamps, floodlights, wall washing lights, color-changing lights, etc. have been widely used in actual projects. These lamps not only involve power supply requirements of different voltage levels in engineering applications, but also rely on various control protocols such as DMX512, TTL, DALI, etc. for signal transmission and parameter setting. At the same time, the testing work of lamps at the engineering construction site is gradually evolving from traditional manual operation to portability, automation and intelligence. Especially in scenarios such as multi-lamp joint control, remote detection, and real-time response evaluation in complex environments, there is an urgent need for a multifunctional lighting test system with strong compatibility, high response accuracy and high integration.

[0003] The existing lighting project testing means generally have the following problems: First, most testing systems have a single function and usually only support single voltage output or fixed control protocols, and cannot meet the needs of mixed testing of multiple types of lamps. During the actual construction process, testers often need to carry multiple different power adapters and signal conversion devices, which not only have cumbersome operations and complex connections, but also are extremely prone to wiring errors and signal distortion. Second, the existing testing systems lack a unified data acquisition and analysis mechanism. Key operation data of lamps during the controlled response process, such as voltage, current, brightness change images and control feedback information, are often unable to be systematically collected and structurally stored, resulting in the inability to quantitatively evaluate the test results. The judgment of project quality still relies on manual experience or subjective observation, and the accuracy and traceability are relatively poor.

[0004] Third, although some testing systems have begun to try to introduce data-driven analysis modules in recent years, they still stay in the stage of simple data recording or threshold comparison for anomaly determination, and lack the ability to deeply model multi-modal time series data. For example, in the face of problems such as time delay, fluctuation, and non-linear coupling between brightness response and electrical signals, the existing systems cannot analyze and reconstruct through a unified model, and it is also difficult to identify potential operation anomalies or device aging risks. In addition, the current lamp testing systems generally do not have an optimized learning algorithm or structural adaptability. Facing different lamp categories or application environments, their model configuration and determination mechanisms still rely on manual adjustment, with a low degree of intelligence and unable to dynamically optimize the model structure and parameters.

[0005] Furthermore, existing systems mostly use fixed rules for anomaly detection, such as fixed voltage fluctuation thresholds, fixed brightness delay ranges, etc. In the face of complex on-site environments, the response characteristics of lamps from different manufacturers, or new intelligent lamps, this method often results in misjudgments or missed detections. At the same time, the processing link after the generation of anomaly detection results in existing systems is also imperfect, often lacking an automated report compilation mechanism, local prompt, and remote upload synchronization mechanism, which affects the ability of engineering personnel to respond in a timely manner and collaborate remotely across regions, and restricts the closed-loop management of the testing process.

[0006] In addition, most existing testing equipment exists in the form of desktop or fixed installation, lacking good mobility and integration. Especially in outdoor, large-scale projects, or situations with tight construction cycles, it cannot be quickly deployed and removed, seriously affecting the testing efficiency and project progress. In some scenarios that require rapid fault review and temporary extended testing, existing systems are difficult to handle operations such as multiple voltage output switching, rapid control protocol loading, and structural reconstruction modeling, restricting their practical value.

[0007] Therefore, how to provide a multi-functional mobile power supply lighting engineering test system and method is an urgent problem for those skilled in the art. Summary of the Invention

[0008] An object of the present invention is to propose a multi-functional mobile power supply lighting engineering test system and method. The present invention fully integrates technologies such as multi-voltage power supply switching, control protocol adaptability, heterogeneous multi-modal data acquisition, variational autoencoder modeling analysis, and beetle antenna search algorithm optimization, and details how to achieve intelligent modeling and anomaly recognition of lamp operation response data during on-site testing, with the advantages of strong compatibility, high intelligence level, accurate anomaly detection, and convenient portable deployment.

[0009] A multi-functional mobile power supply lighting engineering test method according to an embodiment of the present invention includes the following steps:

[0010] S1. Initialize the multi-functional mobile power supply lighting engineering test system, start the self-check program, perform status detection, and confirm that all parts of the device are working properly and in a state to be tested;

[0011] S2. According to the power supply parameters and control protocol type of the lamp to be tested, complete the connection operation between the lamp and the multi-functional mobile power supply lighting engineering test system;

[0012] S3. Activate the power module, select DC5V, DC12V, DC24V, and AC220V power supply channels according to the voltage level required by the lamp, use the energy storage module, and dynamically control the power output parameters through the digital signal processor;

[0013] S4. Activate the lighting control module, select the DMX512 and TTL signal output modes according to the control protocol supported by the connected lamps, set the lighting channel number, control bit width, and transformation mode, and send control signals to the lamps.

[0014] S5. Collect the operation response data of the lamps during the control response process and input it into the variational autoencoder model for feature encoding and decoding.

[0015] S6. Optimize the structural hyperparameters of the variational autoencoder model using the beetle antennae search algorithm, perform parameter optimization through the beetle antennae search algorithm, and generate an optimized variational autoencoder model.

[0016] S7. Compare the reconstructed output of the optimized variational autoencoder model with the operation response data. When the reconstruction error exceeds the preset threshold, determine that the current lamp is abnormal, generate an abnormal report, and perform local display or remote upload to assist testers in positioning analysis and fault repair.

[0017] S8. After the test is completed, turn off the power supply channel and control signal, save all the collected operation response data and analysis reports during this test to the local storage module or synchronize them to the remote platform for engineering file management.

[0018] Optionally, the connection operation between the lamp and the multi-functional mobile power supply lighting engineering test system specifically includes connecting the power input end of the lamp to the corresponding output interface of the power module and connecting the signal input end of the lamp to the control signal output end of the lighting control module.

[0019] Optionally, the S3 specifically includes:

[0020] S31. Obtain the power supply voltage parameter of the lamp to be tested and determine that the target voltage level required by the lamp is one of DC5V, DC12V, DC24V, and AC220V.

[0021] S32. Control the power module to switch to the corresponding output channel according to the determined target voltage level and configure the output circuit matching the target voltage level.

[0022] S33. Activate the energy storage module as the power source for this power supply operation and provide a continuous and stable energy input to the power module.

[0023] S34. Establish a power supply connection path between the energy storage module and the power module so that electrical energy can be efficiently transmitted from the energy storage module to the power module.

[0024] S35. Activate the digital signal processor in the power module and adjust the output voltage and current of the power module in real time according to the target voltage level.

[0025] S36. Output the output voltage dynamically adjusted by the digital signal processor from the power supply module to the connected lamp, complete the power supply process and maintain output stability.

[0026] Optionally, the S4 specifically includes:

[0027] S41. Start the lighting control module, complete the initialization operation of the lighting control module, and prepare the software and hardware resources required for control signal output;

[0028] S42. Detect the control protocol type of the connected lamp, and identify that the protocol supported by the lamp is one of the DMX512 protocol and the TTL protocol;

[0029] S43. According to the identified control protocol type, load the corresponding signal output method in the lighting control module, and establish a control communication path with the lamp;

[0030] S44. According to the lamp model and function requirements, set the number of control channels required by the lamp, and configure the functions corresponding to each channel, such as brightness, color, or mode selection;

[0031] S45. Set the data bit width parameter of the control signal, select the appropriate control data format, and set the mode types required for lighting changes, including constant on, flashing, and gradual change;

[0032] S46. Output the configured control signal from the lighting control module to the control input port of the lamp, and drive the lamp to complete the lighting response action according to the set channels, bit width, and transformation modes.

[0033] Optionally, the S5 specifically includes:

[0034] S51. Collect the operation response data of the lamp during the control response process, including voltage waveform, current waveform, brightness change image, and control signal feedback data, and construct a heterogeneous multi-modal input data set;

[0035] S52. Perform cross-modal encoding processing on the heterogeneous multi-modal input data set, input each modality into the corresponding encoding sub-network for feature extraction, and generate modal feature representations z k , where k is the modality number;

[0036] S53. Introduce a heterogeneous channel dynamic weighting mechanism, set the heterogeneous channel weight sparsity parameter Θ hm , and weight and fuse each modal feature z k to generate a joint latent variable z. The heterogeneous channel weight sparsity parameter represents a structural hyperparameter used to adjust the sparsity of the weight distribution in the fusion of each modal feature;

[0037] S54. Introduce a potential space variable dimension mechanism in the encoder and set the potential dimension adjustable factor Θ dim , control the potential space dimension range to adaptively adjust within a preset interval, and generate a potential representation vector with variable dimensions;

[0038] S55. Input the potential representation vector into the decoder to reconstruct the lamp operation response data and form a model reconstruction output data set

[0039] S56. Adopt a multi-threshold reconstruction regularization mechanism and set the multi-threshold regulation parameter Θ th , according to the type and feature complexity of the operation response data, dynamically configure multiple reconstruction error thresholds, and perform segmented penalty weighting on the reconstruction error based on the multi-threshold regulation parameter;

[0040] S57. Compare the reconstructed output of the optimized variational autoencoder model with the operation response data X, calculate the reconstruction error, and use the above structural hyperparameters Θ hm , Θ dim , Θ th as target variables and input them into the beetle antenna search algorithm for joint optimization.

[0041] Optionally, the specific content of S6 includes:

[0042] S61. Set the set of structural hyperparameters of the variational autoencoder model to be optimized as Θ = {Θ hm , Θ dim , Θ th}, where Θ hm is the heterogeneous channel weight sparsity parameter, Θ dim is the potential dimension adjustable factor, and Θ th is the multi-threshold regulation parameter;

[0043] S62. Initialize the search individuals of the beetle antenna search algorithm, including the initial positions of each beetle antenna individual, the perception direction vector the perception step size the antenna length and allocate a historical direction queue Q i to each beetle antenna individual for recording the historical optimal direction;

[0044] S63. Adopt a dimension-by-dimension perturbation strategy and separately set the perturbation direction d j for each structural hyperparameter dimension to form the direction vector

[0045] S64. Generate left and right perception points on both sides of the current position Θ i :

[0046]

[0047] Among them, Θ L,i represents the position vector of the left perception point generated after offsetting the perception length l in the positive direction along the perturbation direction based on the current position of the i-th search individual, Θ in the positive direction; i Θ R,i represents the position vector of the right perception point generated after offsetting the perception length l in the negative direction along the perturbation direction based on the current position of the i-th search individual. The position vectors of the left and right perception points are respectively input into the variational autoencoder model to obtain the corresponding fitness function values f(Θ ) and f(Θ i ); L,i ) R,i ;

[0048] S65. Define the multi-index fitness function f(Θ) as:

[0049]

[0050] Among them, E rec (Θ) is the reconstruction error of the variational autoencoder model, C model (Θ) is the complexity of the variational autoencoder model, H s (Θ) is the latent space structure entropy, is the reconstruction confidence volatility, and α, β, γ, δ are weight coefficients;

[0051] S66. Adopt a hybrid guiding direction update mechanism to comprehensively generate a new search direction by combining the current perception direction, local trend direction, and historical direction, specifically including:

[0052] Sample multiple perturbation points near the current position Θ i and calculate the pseudo-gradient direction based on the central difference

[0053] Extract the top k optimal directions from the historical direction queue of the individual Apply the exponential decay function to generate the memory weight w m ;

[0054] Fuse the current perception direction the local trend direction with the weighted average of the historical directions to form the update direction vector:

[0055]

[0056] Among them, is the update direction of the i-th individual at the (t + 1)-th iteration, k is the number of historical directions or local perturbation samplings, and λ1, λ2, λ3 are fusion weight coefficients;

[0057] S67. Update the current position according to the fitness difference between the left and right sensing points:

[0058]

[0059] where is the structural hyperparameter vector updated by the i-th search individual after the (t + 1)-th iteration, is the structural hyperparameter vector of the i-th search individual at the t-th iteration, s i is the step size parameter of the i-th search individual, and sign is the sign function;

[0060] S68. Adjust the offset sensing length l i according to the difference Δf i in the fitness values of the left and right sensing points. If Δf i > δ high , then increase l i ; if Δf i < δ low , then decrease l i , where δ high is the upper threshold of the sensing difference, and δ low is the lower threshold of the sensing difference;

[0061] S69. Perform boundary constraint processing on the updated hyperparameter Θ i and write the directions of the individuals with excellent performance into the historical direction queue Q i ;

[0062] S610. Repeat steps S63 to S69 until the maximum number of iterations is reached or the convergence condition is satisfied, and output the optimal hyperparameter combination Θ * and use it to construct an optimized variational autoencoder model.

[0063] Optionally, the specific content of S7 includes:

[0064] S71. Load the variational autoencoder model optimized by the longhorn beetle antenna search algorithm into the multi-functional mobile power supply lighting engineering test system as an intelligent analysis model for running response data;

[0065] S72. Input the collected lamp running response data into the optimized variational autoencoder model to generate the reconstructed output data of the optimized variational autoencoder model;

[0066] S73. Compare the reconstructed output data of the optimized variational autoencoder model with the operation response data to identify the deviation results between the various data items;

[0067] S74. Compare the deviation values in the comparison results with the set error judgment threshold to determine whether the current operating state of the lamp exceeds the preset normal range;

[0068] S75. When the deviation value of any data item exceeds the error judgment threshold, it is determined that the lamp has an abnormal operation, and the multi-functional mobile power supply lighting project test system automatically generates the abnormal type, abnormal location and relevant analysis information;

[0069] S76. Compile the generated abnormal information into an abnormal report, prompt it through the local display interface, or upload the report to the remote platform for the testers to perform fault location, maintenance decision-making and maintenance operations.

[0070] A multi-functional mobile power supply lighting project test system according to an embodiment of the present invention includes the following modules:

[0071] The system initialization module is used to start the multi-functional mobile power supply lighting project test system, perform self-checks and status detections on each functional unit of the device, and confirm that it is in a state to be tested;

[0072] The power supply module is used to select a power supply channel according to the voltage level required by the lamp, and perform real-time dynamic adjustment of the output parameters through a digital signal processor;

[0073] The energy storage module is used to provide a continuous and stable power source for the power supply module to support the operation requirements of the system in a portable or off-grid state;

[0074] The lighting control module is used to select a signal output method according to the control protocol supported by the lamp, set the control channel, bit width and change mode, and send the control signal to the lamp signal input end;

[0075] The data acquisition and processing module is used to collect the operation response data of the lamp during the test process and construct a multi-modal data set for modeling analysis;

[0076] The model modeling and optimization module is used to input the collected data into the variational autoencoder model, and jointly optimize its structural hyperparameters based on the beetle antennae search algorithm to complete feature extraction, reconstruction and structural dynamic adjustment;

[0077] The abnormal recognition and result output module is used to compare the model reconstruction output with the original data, identify the abnormal state when the reconstruction error exceeds the threshold, and automatically generate an abnormal report and display, upload and archive it locally or remotely.

[0078] The beneficial effects of the present invention are:

[0079] The present invention constructs an integrated multi-functional mobile power supply lighting project test system and method, which overcomes various technical defects of existing test equipment, such as single function, complex connection, weak data processing ability, and poor intelligence in abnormal recognition, and significantly improves the efficiency, accuracy, and intelligent level of lighting project testing. The test system of the present invention integrates a power supply module, an energy storage module, and a lighting control module in terms of structure, and can flexibly adapt to different voltage levels (such as DC5V, DC12V, DC24V, AC220V) and control protocols (such as DMX512, TTL) required by various lamps, simplifies the on-site wiring and operation process, and significantly improves the test flexibility and engineering adaptability.

[0080] In terms of the test method, the present invention innovatively introduces a heterogeneous multi-modal operation response data acquisition mechanism, realizes the synchronous acquisition and fusion modeling of key operation data such as voltage waveforms, current waveforms, brightness images, and control signal feedback of lamps under controlled states, and constructs a feature input basis suitable for complex dynamic response scenarios. By constructing a variational autoencoder model with a variable structure, the collected data is encoded, decoded, and reconstructed in multiple layers, and a heterogeneous channel weight sparse control, a latent space dimension adjustable mechanism, and a multi-threshold regularization mechanism are introduced into the model, so that the model has stronger feature extraction ability and discriminant robustness to abnormal states.

[0081] Furthermore, the present invention uses an improved beetle antenna search algorithm to optimize the structural hyperparameters of the variational autoencoder model. By introducing a dimension-by-dimension perturbation strategy, an adaptive perception length adjustment mechanism, and a hybrid guiding direction update mechanism, automatic matching and global optimization of the model structure under different lamp types and response behaviors are realized. The optimized model can achieve high-precision fitting and reconstruction of the operating state, and accurately judge whether there is an abnormal operating condition through error comparison. Once an abnormality is identified, a complete report including the abnormal type, location, and analysis conclusion will be generated, and local display or remote upload is supported.

[0082] Through the above methods and systems, the present invention not only realizes the whole process of intelligence, automation, and closed-loop control from data acquisition to abnormal analysis, but also has high on-site deployability and compatibility with multiple types of lamps, significantly improves the work efficiency, data reliability, and problem response speed in the testing process of engineering lamps, and has good engineering adaptability and industrial promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0084] Figure 1Flow chart of a multi-functional mobile power supply lighting project test method proposed by the present invention;

[0085] Figure 2 Structural schematic diagram of a multi-functional mobile power supply lighting project test system proposed by the present invention. Detailed implementation manners

[0086] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0087] Refer to Figure 1 , a multi-functional mobile power supply lighting project test method, including the following steps:

[0088] S1. Initialize the multi-functional mobile power supply lighting project test system, start the self-check program, perform status detection, and confirm that all parts of the device are working properly and in a state to be tested;

[0089] S2. According to the power supply parameters and control protocol type of the to-be-tested lamp, complete the connection operation between the lamp and the multi-functional mobile power supply lighting project test system;

[0090] S3. Activate the power supply module, select the DC5V, DC12V, DC24V, and AC220V power supply channels according to the voltage level required by the lamp, adopt the energy storage module, and dynamically control the power output parameters through the digital signal processor;

[0091] S4. Start the lighting control module, select the DMX512 and TTL signal output modes according to the control protocol supported by the connected lamp, set the number of lighting channels, control bit width, and transformation mode, and send control signals to the lamp;

[0092] S5. Collect the operation response data of the lamp during the control response process, and input it into the variational autoencoder model for feature encoding and decoding;

[0093] S6. Optimize the structural hyperparameters of the variational autoencoder model by using the beetle antennae search algorithm, perform parameter optimization through the beetle antennae search algorithm, and generate an optimized variational autoencoder model;

[0094] S7. Compare the reconstructed output of the optimized variational autoencoder model with the operation response data. When the reconstruction error exceeds the preset threshold, it is determined that the current lamp is abnormal, and an exception report is generated and locally displayed or remotely uploaded to assist the tester in location analysis and fault repair;

[0095] S8. After the test is completed, turn off the power supply channel and control signal, and save all the collected operation response data and analysis reports during this test to the local storage module or synchronize them to the remote platform for engineering file management.

[0096] In this embodiment, the connection operation between the lamp and the multi-functional mobile power supply lighting engineering test system specifically includes connecting the power input end of the lamp to the corresponding output interface of the power module, and connecting the signal input end of the lamp to the control signal output end of the lighting control module.

[0097] In this embodiment, S3 specifically includes:

[0098] S31. Obtain the power supply voltage parameter of the lamp to be tested, and determine that the target voltage level required by the lamp is one of DC5V, DC12V, DC24V, and AC220V;

[0099] S32. Control the power module to switch to the corresponding output channel according to the determined target voltage level, and configure the output circuit matching the target voltage level;

[0100] S33. Start the energy storage module as the power source for this power supply operation to provide continuous and stable energy input to the power module;

[0101] S34. Establish a power supply connection path between the energy storage module and the power module so that electric energy can be efficiently transmitted from the energy storage module to the power module;

[0102] S35. Activate the digital signal processor in the power module and adjust the output voltage and current of the power module in real time according to the target voltage level;

[0103] S36. Output the output voltage dynamically adjusted by the digital signal processor from the power module to the connected lamp to complete the power supply process and maintain output stability.

[0104] In this embodiment, S4 specifically includes:

[0105] S41. Start the lighting control module, complete the initialization operation of the lighting control module, and prepare the software and hardware resources required for control signal output;

[0106] S42. Detect the control protocol type of the connected lamp and identify that the protocol supported by the lamp is one of the DMX512 protocol and the TTL protocol;

[0107] S43. According to the identified control protocol type, load the corresponding signal output mode in the lighting control module and establish a control communication path with the lamp;

[0108] S44. Set the number of control channels required for the lamp according to the lamp model and functional requirements, and configure the functions corresponding to each channel, such as brightness, color, or mode selection;

[0109] S45. Set the data bit width parameter of the control signal, select an appropriate control data format, and set the mode types required for the light change, including constant on, flashing, and gradual change;

[0110] S46. Output the configured control signal from the light control module to the control input port of the lamp, and drive the lamp to complete the light response action according to the set channels, bit width, and transformation mode.

[0111] In this embodiment, the specific steps of S5 are as follows:

[0112] S51. Collect the operation response data of the lamp during the control response process, including voltage waveform, current waveform, brightness change image, and control signal feedback data, and construct a heterogeneous multi-modal input data set;

[0113] S52. Perform cross-modal encoding processing on the heterogeneous multi-modal input data set, input each modality into the corresponding encoding sub-network for feature extraction, and generate modal feature representations z k , where k is the modality number;

[0114] S53. Introduce a heterogeneous channel dynamic weighting mechanism, set the heterogeneous channel weight sparsity parameter Θ hm , and perform weighted fusion on each modal feature z k to generate a joint latent variable z. The heterogeneous channel weight sparsity parameter represents a structural hyperparameter used to adjust the sparsity of the weight distribution in the fusion of each modal feature;

[0115] S54. Introduce a latent space variable dimension mechanism in the encoder, set the latent dimension adjustable factor Θ dim , and control the latent space dimension range to adaptively adjust within a preset interval to generate a latent representation vector with variable dimensions;

[0116] S55. Input the latent representation vector into the decoder to reconstruct the lamp operation response data and form a model reconstruction output data set

[0117] S56. Adopt a multi-threshold reconstruction regularization mechanism, set the multi-threshold regulation parameter Θ th , dynamically configure multiple reconstruction error thresholds according to the type and feature complexity of the operation response data, and perform segmented penalty weighting on the reconstruction error based on the multi-threshold regulation parameter;

[0118] S57. The reconstructed output of the optimized variational autoencoder model Compare with the running response data X, calculate the reconstruction error, and use the above structural hyperparameter Θ hm , Θ dim , Θ th The target variable is input into the beetle whisker search algorithm for joint optimization.

[0119] In this implementation manner, S6 specifically includes:

[0120] S61, set the variational autoencoder model structure hyperparameter set to be optimized as Θ = {Θ hm ,Θ dim ,Θ th}, where Θ hm is the sparsity parameter of heterogeneous channel weights, Θ dim is the potential dimension adjustable factor, Θ th is a multi-threshold control parameter;

[0121] S62, initializing the search individuals of the beetle whisker search algorithm, including the initial position of each beetle whisker individual Perceived direction vector Perception step length Tentacle length And assign a historical direction queue Q to each individual beetle beetle i , used to record the historical optimal direction;

[0122] S63, using the dimension perturbation strategy, set the perturbation direction d for each structural hyperparameter dimension separately j , which constitutes the direction vector of the individual beetle whiskers

[0123] S64, at the current position Θ i Generate left and right perception points on both sides:

[0124]

[0125] Among them, Θ L,i Indicates that based on the current position of the i-th search individual, along the disturbance direction Positive direction offset perception length l i The left perception point position vector generated later, Θ R,i Indicates that based on the current position of the i-th search individual, along the disturbance direction Negative direction offset perception length l i The right perception point position vector generated after the left perception point position vector and the right perception point position vector are respectively input into the variational autoencoder model to obtain the corresponding fitness function value f(Θ L,i ), f(Θ R,i );

[0126] S65. Define the multi-index fitness function \(f(\Theta)\) as follows:

[0127]

[0128] Among them, \(E\) rec \((\Theta)\) is the reconstruction error of the variational autoencoder model, \(C\) model \((\Theta)\) is the complexity of the variational autoencoder model, \(H\) s \((\Theta)\) is the latent space structure entropy, is the reconstruction confidence volatility, and \(\alpha, \beta, \gamma, \delta\) are weight coefficients;

[0129] S66. Adopt a hybrid guiding direction update mechanism to comprehensively combine the current perception direction, local trend direction, and historical direction to generate a new search direction, specifically including:

[0130] Sample multiple perturbation points near the current position \(\Theta\) i and calculate the pseudo-gradient direction based on central difference

[0131] Extract the top \(k\) optimal directions from the historical direction queue of the individual Apply the exponential decay function to generate the memory weight \(w\) m ;

[0132] Fuse the current perception direction local trend direction with the weighted average of the historical directions to form the update direction vector:

[0133]

[0134] Among them, is the update direction of the \(i\)-th individual at the \((t + 1)\)-th iteration, \(k\) is the number of historical directions or local perturbation samplings, and \(\lambda_1, \lambda_2, \lambda_3\) are fusion weight coefficients;

[0135] S67. Update the current position according to the fitness difference between the left and right perception points:

[0136]

[0137] Among them, is the structural hyperparameter vector updated by the \(i\)-th search individual after the \((t + 1)\)-th iteration, is the structural hyperparameter vector of the \(i\)-th search individual at the \(t\)-th iteration, \(s\) i is the step size parameter of the \(i\)-th search individual, and sign is the sign function;

[0138] S68. Adjust the offset perception length \(l\) according to the difference \(\Delta f\) in the fitness values of the left and right perception points i i, if Δf i > δ high , then increase l i ; if Δf i < δ low , then decrease l i , where δ high is the upper threshold of the perceived difference, and δ low is the lower threshold of the perceived difference;

[0139] S69. Perform boundary constraint processing on the updated hyperparameter Θ i and write the direction of the individuals with excellent performance into the historical direction queue Q i ;

[0140] S610. Repeat steps S63 to S69 until the maximum number of iterations is reached or the convergence condition is satisfied, and output the optimal hyperparameter combination Θ * and use it to construct an optimized variational autoencoder model.

[0141] In this embodiment, S7 specifically includes:

[0142] S71. Load the variational autoencoder model optimized by the longhorn beetle antenna search algorithm into the multi-functional mobile power supply lighting project test system as an intelligent analysis model for operation response data;

[0143] S72. Input the collected lamp operation response data into the optimized variational autoencoder model to generate the reconstructed output data of the optimized variational autoencoder model;

[0144] S73. Compare the reconstructed output data of the optimized variational autoencoder model with the operation response data to identify the deviation results between the data items;

[0145] S74. Compare the deviation value in the comparison result with the set error judgment threshold to determine whether the current lamp operation state exceeds the preset normal range;

[0146] S75. When the deviation value of any data item exceeds the error judgment threshold, it is determined that the lamp has an abnormal operation, and the multi-functional mobile power supply lighting project test system automatically generates the abnormal type, abnormal location and related analysis information;

[0147] S76. Compile the generated abnormal information into an abnormal report, prompt it through the local display interface, or upload the report to the remote platform for the tester to perform fault location, repair decision-making and maintenance operations.

[0148] Reference Figure 2 , a multi-functional mobile power supply lighting project test system, includes the following modules:

[0149] The system initialization module is used to start the multi-functional mobile power supply lighting project test system, perform self-checks and status detections on each functional unit of the device, and confirm that it is in a state to be tested;

[0150] The power supply module is used to select a power supply channel according to the voltage level required by the lamp, and perform real-time dynamic adjustment of the output parameters through a digital signal processor;

[0151] The energy storage module is used to provide a continuous and stable power source for the power supply module, and support the operation requirements of the system in a portable or off-grid state;

[0152] The lighting control module is used to select a signal output mode according to the control protocol supported by the lamp, set the control channel, bit width and change mode, and send the control signal to the lamp signal input terminal;

[0153] The data acquisition and processing module is used to collect the lamp operation response data during the test process, and construct a multi-modal data set for modeling and analysis;

[0154] The model modeling and optimization module is used to input the collected data into the variational autoencoder model, and jointly optimize its structural hyperparameters based on the beetle antennae search algorithm to complete feature extraction, reconstruction and structural dynamic adjustment;

[0155] The anomaly recognition and result output module is used to compare the model reconstruction output with the original data, identify the abnormal state when the reconstruction error exceeds the threshold, automatically generate an anomaly report, and display, upload and archive it locally or remotely.

[0156] Embodiment 1:

[0157] To verify the feasibility of the present invention in implementation, the present invention is applied to a provincial key landscape lighting project. The construction unit plans to conduct power-on tests and functional acceptance for a group of outdoor lamps including wall washing lights, RGB floodlights and intelligent color-changing lights. Due to the large differences in lamp brands, power supply standards and control protocols, there are multiple practical problems at the test site: traditional test equipment cannot support the rapid switching of different voltage levels, control protocol identification requires manual operation, some lamps are slow to respond to signals, there are phenomena such as unqualified brightness, large voltage fluctuations and unstable current, and the original test means lack a unified data acquisition mechanism. Engineering personnel mainly rely on visual observation and experience judgment, resulting in low test efficiency and high misjudgment rate.

[0158] To improve the test accuracy and efficiency, the technical team introduced the multi-functional mobile power supply lighting engineering test system and method described in the present invention. The system is powered by a built-in energy storage module, supports four voltage outputs of DC5V, DC12V, DC24V, and AC220V, has the function of automatically identifying DMX512 and TTL control protocols, is equipped with a lamp response behavior modeling module constructed based on variational autoencoders, and introduces the beetle antenna search algorithm to optimize the model structure, and can automatically complete the analysis and modeling of sampled data, anomaly identification, and report output.

[0159] In the actual application process, the technical personnel deployed the test equipment in the center of the construction area and connected the A01 wall washer, B01 floodlight, and C01 intelligent lamp to the system test channels respectively. After the system completes initialization, it automatically identifies the power supply parameters and control protocols of the lamps through the connection module, accurately switches the voltage channels, and configures the signal output format according to the control protocol. After the test starts, the device real-time collects the voltage waveforms, current changes, brightness images, and feedback signals of the lamps during the controlled operation process to construct multi-modal operation response data.

[0160] Subsequently, the system inputs these data into the optimized variational autoencoder model for reconstruction analysis. Combining the structure hyperparameters optimized by the beetle antenna search algorithm, the system can make adaptive adjustments to the response characteristics of different lamps and accurately model their operating states. During the comparison process between the reconstructed output and the original data, the system identifies that there are obvious anomalies in some lamps before optimization, such as low brightness of A01, poor voltage stability of B01, and large current fluctuations of C01. After optimizing the power supply output and modeling parameters, the response performance of the three lamps has been significantly improved.

[0161] After the system completes the test, it automatically generates a complete anomaly determination report, which details the test time, lamp number, test indicators, deviation conditions, and results after optimization, and uploads it to the project scheduling platform through the remote communication module to support subsequent project filing and quality inspection recheck. The entire test process can be completed by 1 person. The test and optimization process of 3 lamps takes less than 25 minutes in total, and the efficiency is improved by about 65% compared with the traditional method.

[0162] Table 1 Comparison table of the effects of typical lamps before and after testing

[0163]

[0164] In this typical lamp test, we selected three representative types of lamps with numbers (A01, B01, and C01), corresponding to three key test indicators: brightness response, voltage stability, and current stability. During the test, the response data under traditional test methods (i.e., the measured values before optimization) were first collected. Subsequently, the multifunctional mobile power supply lighting engineering test system proposed in the present invention was applied for optimization testing, and the optimized results were collected again and compared and analyzed with the expected values.

[0165] For the wall washer lamp numbered A01, its test indicator is brightness response, and the expected brightness is 800 units. In traditional tests, due to unstable adjustment of the supply voltage and low output control accuracy, the actual response was only 740 units, with problems such as low brightness and slow response. After the system dynamically adjusted the output channels, finely controlled the power module, and combined with the optimized variational autoencoder model for modeling and judgment, the brightness response value was increased to 795 units, approaching the theoretical target. The brightness improvement was 55 units, and the difference was clearly observable to the naked eye. Moreover, the improvement process was stable without stroboscopic effects.

[0166] The lamp numbered B01 is a floodlight, and its concerned indicator is voltage stability. The expected voltage of the system is 220.0V. In the state before optimization, the controlled output voltage was 216.3V, and the voltage deviation reached 3.7V. Voltage fluctuations even occurred in some time periods, posing a long-term loss risk to the lamp drive circuit. In the system of the present invention, the digital signal processor combined with the output capacity of the energy storage module performs dynamic closed-loop control on the voltage, and the response data is transmitted back for modeling for predictive adjustment. After optimization, the output voltage reached 219.2V, with an error of only 0.8V, and the stability was significantly improved.

[0167] The lamp numbered C01 is a color-changing intelligent lamp, and its core problem is current stability. The current target of this lamp during operation is 1.20A. However, under the traditional power supply method, due to slow feedback response and inaccurate model, the current dropped to 1.02A, resulting in not only insufficient brightness but also the phenomenon of not responding to control commands. After optimization by the system of the present invention, the model structure automatically adapts to the lamp type, and the current output is increased to 1.18A, approaching the rated parameters, with the error controlled within 0.02A. Finally, the output data curve is stable and the signal consistency is high.

[0168] Generally speaking, the test data of the three groups of typical lamps fully prove that the system of the present invention has significant advantages in improving the lamp performance, enhancing the operation stability and accuracy. The optimized data show that the brightness deviation is significantly reduced, the voltage is more stable, the current approaches the target value, and the whole optimization process is automatically completed by the system without manual intervention. This intelligent modeling and dynamic optimization mechanism provides strong support for high-efficiency and high-accuracy testing in large-scale lighting engineering projects, and also reflects the comprehensive competitiveness of the present invention in terms of testing accuracy, engineering adaptability and remote response ability.

[0169] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, shall be covered by the protection scope of the present invention.

Claims

1. A method for testing a multi-functional mobile power supply for lighting engineering, characterized in that, It includes the following steps: S1. Initialize the multi-functional mobile power supply lighting project test system, start the self-check program, conduct status detection, and confirm that all parts of the device are working properly and in a state to be tested; S2. According to the power supply parameters and control protocol type of the to-be-tested lamp, complete the connection operation between the lamp and the multi-functional mobile power supply lighting project test system; S3. Activate the power module, select the power supply channels of DC5V, DC12V, DC24V, and AC220V according to the voltage level required by the lamp, adopt the energy storage module, and dynamically control the power output parameters through the digital signal processor; S4. Start the lighting control module, select the DMX512 and TTL signal output modes according to the control protocol supported by the connected lamp, set the number of lighting channels, control bit width, and transformation mode, and send control signals to the lamp; S5. Collect the operation response data of the lamp during the control response process, and input it into the variational autoencoder model for feature encoding and decoding; S6. Optimize the structural hyperparameters of the variational autoencoder model using the beetle antennae search algorithm, perform parameter optimization through the beetle antennae search algorithm, and generate an optimized variational autoencoder model; S7. Compare the reconstructed output of the optimized variational autoencoder model with the operation response data. When the reconstruction error exceeds the preset threshold, determine that the current lamp is abnormal, generate an abnormal report and display it locally or upload it remotely to assist the tester in positioning analysis and fault repair; S8. After the test is completed, turn off the power supply channel and control signal, save all the collected operation response data and analysis report in this test to the local storage module or synchronize it to the remote platform for engineering file management.

2. The multifunctional mobile power supply lighting project testing method according to claim 1, wherein The connection operation between the lamp and the multi-functional mobile power supply lighting project test system specifically includes connecting the power input end of the lamp to the corresponding output interface of the power module, and connecting the signal input end of the lamp to the control signal output end of the lighting control module.

3. A multifunctional mobile power supply lighting project testing method according to claim 1, characterized in that, The specific content of S3 includes: S31. Obtain the power supply voltage parameters of the to-be-tested lamp, and determine that the target voltage level required by the lamp is one of DC5V, DC12V, DC24V, and AC220V; S32. Control the power module to switch to the corresponding output channel according to the determined target voltage level, and configure the output circuit matching the target voltage level; S33. Start the energy storage module as the power source for this power supply operation to provide continuous and stable energy input to the power module; S34. Establish a power supply connection path between the energy storage module and the power module, so that electric energy can be efficiently transmitted from the energy storage module to the power module; S35. Activate the digital signal processor in the power module, and adjust the output voltage and current of the power module in real time according to the target voltage level; S36. Output the output voltage dynamically adjusted by the digital signal processor from the power module to the connected lamp to complete the power supply process and maintain output stability.

4. A multifunctional mobile power supply lighting project testing method according to claim 1, characterized in that The specific content of S4 includes: S41. Start the lighting control module, complete the initialization operation of the lighting control module, and prepare the software and hardware resources required for control signal output; S42. Detect the control protocol type of the connected lamp, and identify that the protocol supported by the lamp is one of the DMX512 protocol and the TTL protocol; S43. According to the identified control protocol type, load the corresponding signal output method in the lighting control module, and establish a control communication path with the lamp; S44. According to the lamp model and functional requirements, set the number of control channels required by the lamp, and configure the functions corresponding to each channel, such as brightness, color, or mode selection; S45. Set the data bit width parameter of the control signal, select the appropriate control data format, and set the mode types required for lighting changes, including constant on, flashing, and fading; S46. Output the configured control signal from the lighting control module to the control input port of the lamp, and drive the lamp to complete the lighting response action according to the set channels, bit width, and transformation modes.

5. A method for testing a lighting project of a multifunctional mobile power supply according to claim 1, characterized in that, The specific steps of S5 are as follows: S51. Collect the operation response data of the lamp during the control response process, including voltage waveforms, current waveforms, brightness change images, and control signal feedback data, and construct a heterogeneous multi-modal input data set; S52. Perform cross-modal encoding processing on the heterogeneous multi-modal input data set, input each modality into the corresponding encoding sub-network respectively for feature extraction, and generate modal feature representations \(z^{(k)}\), where \(k\) is the modality number; k , where \(k\) is the modality number; S53. Introduce a heterogeneous channel dynamic weighting mechanism and set the sparsity parameter Θ of the heterogeneous channel weights hm , and weight and fuse each modal feature z k to generate a joint latent variable z. The sparsity parameter of the heterogeneous channel weights represents a structural hyperparameter used to adjust the sparsity of the weight distribution in the fusion of each modal feature; S54. Introduce a potential space variable dimension mechanism in the encoder and set a potential dimension adjustable factor Θ dim , and control the potential space dimension range to adaptively adjust within a preset interval to generate a potential representation vector with variable dimensions; S55. Input the latent representation vector into the decoder to reconstruct the lamp operation response data, forming a model reconstruction output data set S56. Adopt a multi-threshold reconstruction regularization mechanism and set a multi-threshold regulation parameter Θ th , dynamically configure multiple reconstruction error thresholds according to the type and feature complexity of the operation response data, and perform piecewise penalty weighting on the reconstruction error based on the multi-threshold regulation parameter; S57. Reconstruct and output the optimized variational autoencoder model Compare it with the operation response data X, calculate the reconstruction error, and use the above structural hyperparameters Θ hm , Θ dim , Θ th as the target variables and input them into the beetle antenna search algorithm for joint optimization.

6. A multifunctional mobile power supply lighting project testing method according to claim 1, characterized in that The specific steps of S6 are as follows: S61. Set the hyperparameter set of the variational autoencoder model structure to be optimized as Θ = {Θ hm , Θ dim , Θ th}, where Θ hm is the heterogeneous channel weight sparsity parameter, Θ dim is the latent dimension adjustable factor, and Θ th is the multi-threshold regulation parameter; S62. Initialize the search individuals of the beetle antenna search algorithm, including the initial positions of each beetle antenna individual Perception direction vector Perception step size Antenna length And allocate a historical direction queue Q for each beetle antenna individual i , which is used to record the historical optimal direction; S63. Adopt a dimension - by - dimension perturbation strategy, and set a perturbation direction d for each dimension of the structural hyperparameter separately j , forming the direction vector of the beetle antenna individual S64. At the current position Θ i Generate left and right sensing points on both sides respectively: Among them, Θ L,i represents the position vector of the left sensing point generated after offsetting the sensing length l in the positive direction along the perturbation direction based on the current position of the i-th search individual ; Θ i represents the position vector of the right sensing point generated after offsetting the sensing length l in the negative direction along the perturbation direction based on the current position of the i-th search individual. The position vectors of the left and right sensing points are respectively input into the variational autoencoder model to obtain the corresponding fitness function values f(Θ R,i ), f(Θ ); i After that, the left sensing point position vector and the right sensing point position vector are respectively input into the variational autoencoder model to obtain the corresponding fitness function values f(Θ L,i ), f(Θ R,i ); S65. Define the multi-index fitness function f(Θ) as: Among them, E rec (Θ) is the reconstruction error of the variational autoencoder model, C model (Θ) is the complexity of the variational autoencoder model, H s (Θ) is the latent space structure entropy, is the reconstruction confidence volatility, and α, β, γ, δ are weight coefficients; S66. Adopt a hybrid guiding direction update mechanism to comprehensively consider the current perception direction, local trend direction, and historical direction to generate a new search direction. Specifically, it includes: At the current position Θ i Sample multiple perturbation points near it and calculate the pseudo-gradient direction based on central difference Extract the top k optimal directions from the historical direction queue of an individual Apply an exponential decay function to generate the memory weight w m ; The current perceived direction Local trend direction is fused with the weighted average of the historical directions to form an updated direction vector: Among them, is the update direction of the i-th individual at the (t + 1)-th iteration, k is the number of historical directions or local perturbation samplings, and λ1, λ2, and λ3 are fusion weight coefficients; S67. Update the current position according to the fitness difference between the left and right perception points; Among them, is the structural hyperparameter vector updated after the (t + 1)-th iteration of the i-th search individual, is the structural hyperparameter vector of the i-th search individual at the t-th iteration, and s i is the step size parameter of the i-th search individual, and sign is the sign function; S68. Adjust the offset sensing length l according to the difference Δf between the left and right sensing point fitness values i If Δf i > δ i high , then increase l i ; if Δf i < δ low i , then decrease l, where δ high is the upper threshold of the sensing difference, and δ low is the lower threshold of the sensing difference;​ S69. Boundary constraint processing is performed on the updated hyperparameter Θ i and the direction of the individual with excellent performance is written into the historical direction queue Q i ; S610. Repeat steps S63 to S69 until the maximum number of iterations is reached or the convergence condition is satisfied, and output the optimal hyperparameter combination Θ * , and use it to construct an optimized variational autoencoder model.

7. A multifunctional mobile power supply lighting project testing method according to claim 1, characterized in that The specific steps of S7 are as follows: S71. Load the variational autoencoder model optimized by the beetle antennae search algorithm into the multi-functional mobile power supply lighting engineering test system as an intelligent analysis model for operation response data; S72. Input the collected lamp operation response data into the optimized variational autoencoder model to generate the reconstructed output data of the optimized variational autoencoder model; S73. Compare the reconstructed output data of the optimized variational autoencoder model with the operation response data to identify the deviation results between the data items; S74. Compare the deviation value in the comparison result with the set error judgment threshold to determine whether the current lamp operation state exceeds the preset normal range; S75. When the deviation value of any data item exceeds the error judgment threshold, it is determined that the lamp has an operation abnormality, and the multi-functional mobile power supply lighting engineering test system automatically generates the abnormality type, abnormality location, and related analysis information; S76. Compile the generated abnormality information into an abnormality report, and prompt it through the local display interface, or upload the report to the remote platform for testers to perform fault location, maintenance decision-making, and maintenance operations.

8. A multi-functional mobile power supply lighting project test system, a multi-functional mobile power supply lighting project test method according to any one of claims 1 to 7, characterized in that, It includes the following modules: System initialization module, which is used to start the multi-functional mobile power supply lighting engineering test system, perform self-checks and status detections on each functional unit of the device, and confirm that it is in the state to be tested; Power supply module, which is used to select the power supply channel according to the voltage level required by the lamp, and perform real-time dynamic adjustment of the output parameters through a digital signal processor; Energy storage module, which is used to provide a continuous and stable power source for the power supply module, and support the operation requirements of the system in the portable or off-grid state; The lighting control module is used to select the signal output mode according to the control protocol supported by the lamp, set the control channel, bit width and change mode, and send the control signal to the lamp signal input end; The data acquisition and processing module is used to collect the lamp operation response data during the test and construct a multi-modal data set for modeling and analysis; The model modeling and optimization module is used to input the collected data into the variational autoencoder model, and jointly optimize its structural hyperparameters based on the beetle antennae search algorithm to complete feature extraction, reconstruction and structural dynamic adjustment; The anomaly recognition and result output module is used to compare the model reconstruction output with the original data, identify the abnormal state when the reconstruction error exceeds the threshold, automatically generate an anomaly report and display, upload and archive it locally or remotely.

Images