Building operation and maintenance method and equipment based on OCR (Optical Character Recognition) and energy efficiency analysis, and medium

Abstract

The invention discloses a building operation and maintenance method and equipment based on OCR (Optical Character Recognition) and energy efficiency analysis and a medium, and belongs to the technical field of equipment information recognition and processing. The method comprises the steps of obtaining a power equipment nameplate image; preprocessing the power equipment nameplate image to generate preprocessed image data; processing the preprocessed image data based on a preset deep learning OCR model to generate original text data; performing structured analysis on the original text data to extract key parameters of equipment and generate structured data; calculating the energy consumption level of the equipment based on the structured data, and performing energy efficiency evaluation in combination with preset historical operation data and an industry energy efficiency standard to generate an energy consumption evaluation result; and generating an energy-saving optimization suggestion report according to the energy consumption evaluation result. According to the method, the technical effects of high-precision real-time identification of the nameplate of the power equipment in a complex environment and generation of an executable energy-saving optimization scheme are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of device information recognition and processing, and in particular to a building operation and maintenance method based on OCR recognition and energy efficiency analysis, a device and a medium. BACKGROUND

[0002] Optical character recognition (OCR) technology, as a key means of converting image text into editable digital information, has been widely used in document processing, bill management and industrial device information extraction. In the field of building operation and maintenance, traditional methods rely on manual on-site collection of power equipment nameplate parameters, and energy consumption analysis is performed by manually entering the system. In recent years, some research has attempted to apply general OCR technology to device information collection, using basic image processing and machine learning algorithms to recognize text, and then combining with energy consumption databases for simple comparison and analysis. These technologies have reduced the workload of manual entry to some extent, but are still limited to text recognition in ideal environments and do not fully consider the complex lighting conditions and device diversity characteristics of the building site. Existing technical solutions usually process the recognition and analysis stages independently, and lack an end-to-end automated workflow design.

[0003] The current technical solution has significant limitations: first, in the image recognition stage, traditional OCR methods have poor adaptability to complex environments such as natural light changes, adverse weather interference (such as image blurring caused by rain and fog), and nameplate contamination. In particular in the field of power equipment, problems such as special fonts, metal reflections and inclination angles result in high error rates in character segmentation. Secondly, in the data processing stage, existing methods lack effective structured analysis mechanisms and cannot automatically associate key fields such as voltage levels and power parameters, still requiring manual verification and information completion, significantly reducing processing efficiency. Finally, in terms of energy efficiency analysis, most systems only implement basic energy consumption calculations and fail to combine historical operation data and industry standards for multi-dimensional evaluation, and lack automated report generation capabilities.

[0004] Therefore, how to achieve high-precision real-time recognition of power equipment nameplates in complex environments and automatically generate executable energy-saving optimization schemes has become a technical problem to be solved. SUMMARY

[0005] The embodiments of the present application provide a building operation and maintenance method based on OCR recognition and energy efficiency analysis, a device and a medium, to solve the technical problem of how to achieve high-precision real-time recognition of power equipment nameplates in complex environments and automatically generate executable energy-saving optimization schemes.

[0006] In a first aspect, the embodiments of the present application provide a building operation and maintenance method based on OCR recognition and energy efficiency analysis. The method comprises: acquiring a power equipment nameplate image through an image acquisition device; wherein the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; preprocessing the power equipment nameplate image to perform grayscale, binarization, contrast enhancement and blur removal, and generating preprocessed image data; processing the preprocessed image data based on a preset deep learning OCR model to recognize the text content in the image and generate original text data; structurally analyzing the original text data to extract equipment key parameters and generate structured data; wherein the structured data includes voltage level, power parameter and manufacturer information; calculating the equipment energy consumption level based on the structured data, and combining the preset historical operation data and industry energy efficiency standard to perform energy efficiency evaluation and generate energy consumption evaluation results; generating an energy-saving optimization suggestion report according to the energy consumption evaluation results; wherein the report includes equipment replacement, operation strategy adjustment or energy-saving model recommendation.

[0007] In an implementation manner of the present application, the power equipment nameplate image is preprocessed to perform grayscale, binarization, contrast enhancement and blur removal, and generate preprocessed image data, specifically including: processing the power equipment nameplate image based on a preset grayscale conversion algorithm to convert the color image into grayscale image data; processing the grayscale image data based on a preset histogram equalization algorithm to enhance the image contrast and generate enhanced image data; processing the enhanced image data based on a preset adaptive threshold segmentation algorithm to perform binarization operation and generate binary image data; processing the binary image data based on a preset Gaussian filtering algorithm to remove image noise and blur and generate clear image data; correcting the inclination angle of the clear image data to generate preprocessed image data.

[0008] In an implementation manner of the present application, the grayscale image data is processed based on a preset histogram equalization algorithm to enhance the image contrast and generate enhanced image data, specifically including: processing the grayscale image data based on global histogram equalization to adjust the overall brightness distribution and generate equalized image data; processing the equalized image data based on adaptive histogram equalization to enhance the local area in zones and generate adaptive image data; processing the adaptive image data based on a gamma correction algorithm to optimize the brightness curve and generate corrected image data; processing the corrected image data based on a sharpening filter to enhance the text edge and generate enhanced image data.

[0009] In an implementation manner of the present application, the binary image data is processed based on a preset Gaussian filtering algorithm to remove image noise and blur and generate clear image data, specifically including: processing the binary image data based on a Gaussian kernel construction operation, defining a filter window size, and generating kernel parameter data; processing the binary image data and the kernel parameter data based on a convolution operation, smoothing pixel values, and generating smoothed image data; processing the smoothed image data based on an edge detection algorithm, identifying text contours, and generating edge map data; and processing the edge map data based on a noise suppression operation, removing isolated noise points, and generating clear image data.

[0010] In an implementation manner of the present application, the preprocessed image data is processed based on a preset deep learning OCR model to identify text content in the image and generate original text data, specifically including: processing the preprocessed image data based on a convolutional neural network, extracting character feature vectors, and generating feature map data; processing the feature map data based on a recurrent neural network, modeling text sequence dependency relationships, and generating sequence prediction data; processing the sequence prediction data based on an attention mechanism, weighting and focusing on key text regions, and generating corrected sequence data; processing the corrected sequence data based on a character segmentation algorithm, separating independent character images, and generating character image data; and processing the character image data based on a normalization operation, unifying sizes and directions, and generating standardized character data to determine the original text data.

[0011] In an implementation manner of the present application, the original text data is structurally parsed to extract device key parameters and generate structured data, specifically including: processing the original text data based on a keyword positioning algorithm, identifying pre-defined field keywords, and generating keyword position data; processing the original text data based on a regular expression matching algorithm, extracting parameter values, and generating parameter value data; processing the keyword position data and the parameter value data based on a rule engine, associating fields and values, and generating preliminary structured data; and processing the preliminary structured data based on a template mapping operation, filling a pre-defined database template, and generating structured data.

[0012] In an implementation manner of the present application, the energy consumption level of the device is calculated based on the structured data, and energy efficiency is evaluated in combination with preset historical operation data and industry energy efficiency standards to generate an energy consumption evaluation result, specifically including: processing the structured data based on an energy consumption modeling algorithm to construct a device energy consumption model and generate model parameter data; processing database historical operation data based on a historical data query operation to obtain device operation efficiency indicators and generate historical efficiency data; processing the model parameter data and the historical efficiency data based on an energy efficiency standard comparison algorithm to compare with industry thresholds and generate energy efficiency gap data; and processing the energy efficiency gap data based on a multi-dimensional analysis operation to generate an energy consumption evaluation result by comprehensively considering environmental factors and load conditions.

[0013] In an implementation form of the present application, the energy-saving optimization suggestion report is generated according to the energy consumption evaluation result; specifically, the energy consumption evaluation result is processed based on expert knowledge base query operation, an energy-saving scheme template is matched, and initial suggestion data is generated; the initial suggestion data and historical optimization records are processed based on a machine learning algorithm, the effectiveness of the suggestion is predicted, and optimization probability data is generated; the optimization probability data is processed based on a report generation engine, text and charts are integrated, and report draft data is generated; the report draft data is processed based on automatic checking operation, data consistency is verified, and the energy-saving optimization suggestion is generated.

[0014] In a second aspect, the embodiments of the present application also provide a building operation and maintenance device based on OCR identification and energy efficiency analysis. The device comprises at least one processor and a memory in communication connection with the at least one processor. The memory stores instructions executable by the at least one processor. The instructions are executed by the at least one processor to enable the at least one processor to: acquire a power equipment nameplate image through an image acquisition device; the power equipment nameplate image is taken by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; pre-process the image to perform grayscale, binarization, contrast enhancement and blur removal, and generate pre-processed image data; process the pre-processed image data based on a preset deep learning OCR model to identify the text content in the image and generate original text data; structurally analyze the original text data to extract equipment key parameters and generate structured data; the structured data includes voltage level, power parameters and manufacturer information; calculate the energy consumption level of the equipment based on the structured data, and perform energy efficiency evaluation in combination with preset historical operation data and industry energy efficiency standards to generate an energy consumption evaluation result; generate an energy-saving optimization suggestion report according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment or energy-saving model recommendation.

[0015] In a third aspect, the embodiments of the present application also provide a non-volatile computer storage medium for building operation and maintenance based on OCR recognition and energy efficiency analysis, which stores computer executable instructions, and the computer executable instructions are configured to: acquire a power equipment nameplate image through an image acquisition device; the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; pre-process the image to perform grayscale, binarization, contrast enhancement and blur removal, and generate pre-processed image data; process the pre-processed image data based on a preset deep learning OCR model to identify the text content in the image and generate original text data; structure analysis is performed on the original text data to extract equipment key parameters and generate structured data; the structured data includes voltage level, power parameter and manufacturer information; the energy consumption level of the equipment is calculated based on the structured data, and the energy efficiency is evaluated in combination with the preset historical operation data and the industry energy efficiency standard to generate an energy consumption evaluation result; an energy saving optimization suggestion report is generated according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment or energy saving model recommendation.

[0016] The method, device and medium for building operation and maintenance based on OCR recognition and energy efficiency analysis provided by the embodiments of the present application at least have the following technical effects: The histogram equalization and gamma correction optimize the brightness curve, enhance the contrast between the text and the background, and effectively deal with the blur problem caused by natural light or adverse weather (such as rain and fog). This to some extent guarantees the stability and reliability of the input image, and lays a foundation for subsequent OCR recognition. In the OCR recognition link, the CRNN model and the attention mechanism are integrated, the convolutional neural network is used to extract features, the recurrent neural network is used to model sequence dependence, and the key text area is weighted and focused. This improves the recognition accuracy, enhances the robustness of the system to complex layout and non-standard fonts, and reduces the risk of manual input errors.

[0017] The energy consumption modeling algorithm is used to combine the historical operation data and the industry energy efficiency standard for multi-dimensional analysis, which enhances the accuracy and practicality of the energy consumption evaluation. By querying the historical operation indicators in the database, the historical efficiency data are generated, and the energy efficiency gap is calculated by using the energy efficiency standard comparison algorithm, which to some extent guarantees the consistency of the evaluation result with the actual situation. The expert knowledge base is used to match the energy saving scheme template and the machine learning algorithm to predict the effectiveness, and the optimized probability data are integrated into a standardized report.

[0018] By introducing a multi-task joint loss function and a data enhancement strategy, the system coordinates the text classification and positioning tasks during training, enhancing the model's generalization performance and making it suitable for various scenarios such as power distribution network engineering acceptance and building energy audit. To some extent, this ensures the sustainability of building facility energy management. The overall effect is to improve the efficiency of equipment information collection and enhance the accuracy of data processing. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and serve to explain the principles of the application, and do not limit the application. In the drawings: Figure 1 A building operation and maintenance method based on OCR recognition and energy efficiency analysis is provided for the embodiments of the application; Figure 2 A building operation and maintenance device internal structure schematic diagram based on OCR recognition and energy efficiency analysis is provided for the embodiments of the application. DETAILED DESCRIPTION

[0020] To make the purpose, technical solutions and advantages of the application clearer, the technical solutions of the application will be described clearly and completely below in conjunction with the specific embodiments of the application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0021] The embodiments of the application provide a building operation and maintenance method, device and medium based on OCR recognition and energy efficiency analysis, to solve the technical problem of how to realize high-precision real-time recognition of power equipment nameplates in complex environments and automatically generate executable energy-saving optimization schemes.

[0022] The technical solutions of the embodiments of the application will be described in detail below with reference to the drawings.

[0023] Figure 1 A building operation and maintenance flowchart based on OCR recognition and energy efficiency analysis is provided for the embodiments of the application. As shown in Figure 1 A building operation and maintenance method based on OCR recognition and energy efficiency analysis is provided for the embodiments of the application, which specifically includes the following steps: Step 1, obtaining a power equipment nameplate image through an image acquisition device; wherein the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and device identification information.

[0024] Capturing visual data of power equipment nameplates using image capture devices such as cameras or drones. Nameplate text regions refer to the visible parts of the equipment that contain textual information such as model numbers, parameters, and device identification information includes unique codes or serial numbers.

[0025] In one specific example, A company's maintenance personnel use a drone to automatically capture the nameplate of a distribution transformer in the rain. The drone is equipped with a high-resolution camera that can capture images in low-light conditions, ensuring that the nameplate text region (such as power parameter text) and device identification information (such as serial number code) are fully included. After image acquisition, the image is directly transmitted to the system processing module, avoiding the risk and delay of manual on-site inspection, and improving overall efficiency.

[0026] Step 2, pre-processing power equipment nameplate images for grayscale, binarization, contrast enhancement, and blur removal to generate pre-processed image data.

[0027] The original image is subjected to preliminary optimization processing, including grayscale conversion (converting color images to grayscale images), binarization (converting grayscale images to black and white binary images), contrast enhancement (optimizing brightness differences), and blur removal (reducing noise), ultimately generating high-quality pre-processed image data for subsequent OCR recognition.

[0028] Step 2.1, based on the preset grayscale conversion algorithm to process power equipment nameplate images to convert color images to grayscale image data.

[0029] Using a grayscale conversion algorithm (such as weighted average method) to convert RGB color images to grayscale image data, each pixel only retains brightness information (from 0-255 grayscale values).

[0030] The preset algorithm ensures uniform brightness distribution, avoiding color channel interference in subsequent processing. Grayscale conversion is based on a standard formula (such as Y = 0.299R + 0.587G + 0.114B) to generate grayscale image data, reducing computational complexity.

[0031] In one specific example, after the system receives a color nameplate image of a building HVAC equipment from B, it applies a grayscale conversion algorithm (based on preset weights) to generate a grayscale image. This image only shows light and dark changes, simplifying the positioning of the text region, making subsequent contrast enhancement more efficient.

[0032] Step 2.2, based on the preset histogram equalization algorithm to process grayscale image data to enhance image contrast and generate enhanced image data.

[0033] Adjust the image brightness distribution through the histogram equalization algorithm to make the dark and bright areas more uniform, improving the overall contrast and facilitating text recognition.

[0034] Step 2.2.1, Process the grayscale image data based on global histogram equalization to adjust the overall brightness distribution and generate equalized image data.

[0035] Apply the equalization algorithm to the entire image to stretch the brightness range and make the pixel distribution more uniform.

[0036] This implementation ensures that the text area does not lose details and improves contrast by redistributing pixel values.

[0037] In a specific example, the grayscale image is overall dark, and the system applies the global equalization algorithm to stretch the brightness and generate equalized image data, making the plaque text more recognizable in overcast conditions.

[0038] Step 2.2.2, Process the equalized image data based on adaptive histogram equalization to enhance local areas in partitions and generate adaptive image data.

[0039] Divide the image into small blocks (such as a grid) and independently equalize each region to avoid local distortion caused by global adjustment.

[0040] Apply CLAHE (Contrast Limited Adaptive Histogram Equalization) to process local areas (refer to "enhance local areas in partitions").

[0041] In a specific example, the equalized image has uneven corner brightness, and the system processes it in partitions (such as dividing the image into a 4x4 grid) and independently enhances each block to generate adaptive image data, solving the local shadow problem.

[0042] Step 2.2.3, Process the adaptive image data based on gamma correction algorithm to optimize the brightness curve and generate corrected image data.

[0043] Optimize the image light balance by adjusting the brightness curve non-linearly to make the text clearer.

[0044] Gamma correction is an enhancement technique (refer to "gamma correction"). This implementation combines the previous steps and presets the gamma value (such as 1.8) to adjust the curve, ensuring that the brightness is suitable for OCR input.

[0045] In a specific example, the adaptive image brightness curve is not smooth, and the system applies gamma correction (based on the preset value of 1.5) to optimize the curve and generate corrected image data, reducing overexposed or overdark areas.

[0046] Step 2.2.4, Process the corrected image data based on sharpening filter to enhance text edges and generate enhanced image data.

[0047] Apply a filter (such as Laplacian operator) to highlight text edges and reduce blurring.

[0048] Emphasize sharpening for enhancing text edges (quote "enhancing text edges"). This implementation generates high-quality enhanced image data, laying the foundation for binarization.

[0049] In a specific example, the system corrects the blurring of image text edges using a Laplacian filter sharpening process, generating enhanced image data that improves the clarity of character outlines.

[0050] Step 2.3, process the enhanced image data based on a preset adaptive threshold segmentation algorithm to perform binarization, generating binary image data.

[0051] Use an algorithm to dynamically calculate the threshold value of each region in the image, converting the grayscale image into a black-and-white binary image. The "threshold" refers to the luminance threshold that distinguishes text (black) from background (white); "adaptive" means that the algorithm automatically adjusts the threshold based on local pixel characteristics to deal with uneven lighting.

[0052] Binarization is used to enhance feature contrast (quote "binarization to enhance feature contrast and simplify positioning difficulty"). This implementation presets an algorithm (such as the Otsu method or local threshold method) to process the enhanced image data (from Step 2.2), ensuring that the text area is clearly separated under varying lighting conditions (such as overcast or shadows). This operation simplifies subsequent character segmentation, improves OCR recognition efficiency, and reduces manual intervention.

[0053] In a specific example, the system processes the enhanced image of the B building power distribution cabinet (with local brightness differences). The adaptive threshold segmentation algorithm analyzes image blocks (such as 10x10 pixel regions) and dynamically sets the threshold: lower the threshold in dark areas to preserve text details, and raise the threshold in bright areas to avoid background interference, ultimately generating high-quality binary image data with clear and easily identifiable character outlines.

[0054] Step 2.4, process the binary image data based on a preset Gaussian filter algorithm to remove image noise and blur, generating clear image data.

[0055] Apply a Gaussian filter to smooth the image, eliminating noise (such as salt and pepper noise) and slight blur. The "Gaussian filter" refers to a smoothing algorithm based on the normal distribution principle; "denoising" refers to removing non-text interference points to ensure sharp text edges.

[0056] Gaussian filtering is used to suppress image noise (reference "Gaussian filtering is used to deblur and reduce the impact of noise on recognition results"). The implementation processes binary image data (from step 2.3) through four steps of operations (sub-steps 2.4.1-2.4.4) of building a Gaussian kernel, performing convolution, edge detection, and noise suppression, effectively dealing with image degradation caused by adverse weather (such as raindrops or dust). Finally, clear image data is generated, providing clean input for skew correction and OCR recognition, improving system robustness.

[0057] Step 2.4.1, based on Gaussian kernel construction operation to process binary image data, define filter window size, generate kernel parameter data.

[0058] Define the core parameters of the Gaussian filter, including window size (such as kernel size) and standard deviation (control smoothing strength). Among them, "kernel" refers to the weight matrix used for convolution calculation.

[0059] This implementation presets the window size (such as 5x5 pixel kernel) according to the image size, generates kernel parameter data (including standard deviation 1.0), ensures moderate smoothing effect, and avoids text edge blurring. This step provides basic parameters for subsequent convolution, improving processing efficiency.

[0060] In a specific example, the binary image size is 800x600 pixels, and the system presets the Gaussian kernel window size to 7x7 (standard deviation 1.2), generating kernel parameter data. This data adapts to the image size, preventing over-smoothing that causes character distortion.

[0061] Step 2.4.2, based on convolution operation to process binary image data and kernel parameter data, smooth pixel values, generate smoothed image data.

[0062] Apply Gaussian kernel to image convolution operation, average pixel values to reduce noise. Among them, "convolution" refers to the weighted calculation of kernel matrix and image pixels.

[0063] This implementation processes binary image data and kernel parameter data, generates smoothed image data through weighted average of each pixel. This operation effectively eliminates isolated noise points (such as raindrops), while preserving text structure, ensuring edge detection accuracy.

[0064] In a specific example, the system applies kernel parameter data to binary image, performs convolution operation: each pixel value is based on weighted average of neighborhood pixels (such as 3x3 region). The result generates smoothed image data, significantly reducing random salt and pepper noise in the image.

[0065] Step 2.4.3, based on edge detection algorithm to process smoothed image data, identify text contour, generate edge map data.

[0066] Use an algorithm (e.g. Canny operator) to identify text boundaries, generating contour map data. Here, "edge" refers to the intersection of a character and its background; "detect" specifies the position of the contour line.

[0067] Smooth the image data (from Step 2.4.2) and set the Canny algorithm parameters (e.g. high / low threshold), identify the text contour, and generate edge map data. This step highlights the character shapes, facilitating subsequent noise suppression and improving recognition rates in foggy conditions.

[0068] In a specific example, the system identifies the contour line of the plaque text (e.g. "220V") using the Canny operator, generating edge map data. This data clearly marks the boundaries of the characters, addressing the issue of edge blurring caused by foggy weather.

[0069] Step 2.4.4, Process the edge map data based on noise suppression operations to remove isolated noise points, generating clear image data.

[0070] Filter non-text interference points (e.g. isolated pixels) in the edge map, generating pure image data. Here, "noise points" refer to artifacts that do not belong to the text contour.

[0071] This implementation processes the edge map data (from Step 2.4.3) and applies morphological operations (e.g. opening operation) to remove isolated points, generating final clear image data. This step ensures the integrity of the text region, reduces OCR misrecognition, and is suitable for complex industrial environments.

[0072] In a specific example, the edge map contains non-text noise points (e.g. dust spots), and the system uses morphological opening operation to filter: remove isolated points with an area less than 5 pixels, generating clear image data, making the text region clean and free of interference.

[0073] Step 2.5, Correct the inclination angle of the clear image data, generating preprocessed image data.

[0074] Adjust the image angle to make the plaque text horizontally aligned. Here, "inclination angle" refers to the text skew caused by shooting; "correct" refers to rotating the image to the standard position.

[0075] This implementation processes the clear image data (from Step 2.4) and calculates the inclination angle based on the Hough transform or baseline detection algorithm, automatically rotating the image to generate preprocessed image data. This operation ensures that the text region is horizontal, simplifies subsequent OCR character segmentation, and improves the overall system accuracy.

[0076] In a specific example, the clear image is tilted 15 degrees due to the drone's downward shooting, and the system detects the text baseline and rotates it to the horizontal position, generating standardized preprocessed image data.

[0077] Step 3. Process the pre-processed image data based on a pre-set deep learning OCR model to recognize the text content in the image and generate raw text data.

[0078] Step 3.1. Process the pre-processed image data based on a convolutional neural network to extract character feature vectors and generate feature map data.

[0079] Use a convolutional neural network (CNN) to analyze local image features such as edges and textures, outputting high-dimensional feature vector representation of feature map data.

[0080] Pre-process image data, pre-set multi-layer convolutional layers (such as VGG architecture), learn character stroke and shape features, and generate feature map data. This step preserves spatial information and lays the foundation for sequence modeling, addressing the problem of plaque font diversity.

[0081] In a specific example, the pre-processed image is input into the CNN model, and the system extracts key edge features of the plaque characters (such as "Voltage 220V"), generating feature map data to ensure that subsequent processing focuses on the text area.

[0082] Step 3.2. Process the feature map data based on a recurrent neural network to model text sequence dependencies and generate sequence prediction data.

[0083] Apply a recurrent neural network (RNN) to process feature sequences to capture inter-character context relationships (such as previous and subsequent character dependencies), outputting sequence prediction data.

[0084] The document invention emphasizes the use of RNN for sequence modeling (referring to "RNN for sequence modeling"), and this implementation processes feature map data (from step 3.1) using LSTM or GRU units (long short-term memory network) to model text continuity (such as Chinese character word formation rules) and generate sequence prediction data. This operation improves the robustness of long text recognition and reduces misrecognition caused by segmentation errors.

[0085] In a specific example, the feature map data is input into the RNN layer, and the system models the sequence dependency of "Manufacturer: B Company" to generate sequence prediction data, accurately processing the logical relationship between Chinese word groups.

[0086] Step 3.3. Process the sequence prediction data based on an attention mechanism to weight and focus on key text areas, generating corrected sequence data.

[0087] Use the attention mechanism to dynamically weight important text parts and ignore irrelevant areas, outputting corrected sequence data.

[0088] Process sequence prediction data (from step 3.2), preset weight distribution strategy, focus on core parameters (such as power value), generate corrected sequence data. This step should address the problem of messy nameplate layout and improve recognition accuracy.

[0089] In a specific example, sequence prediction data has noise interference, attention mechanism weighted focus on "power 50kW" key text, generate corrected sequence data, reduce misidentification of background elements.

[0090] Step 3.4, process corrected sequence data based on character segmentation algorithm, separate independent character images, generate character image data.

[0091] Split the text sequence into single character images to facilitate independent recognition and output character image data. "Segmentation" refers to dividing boundaries; "character image" refers to the independent area of each character.

[0092] Use projection method or connected component analysis to separate characters (such as Chinese characters or numbers) and generate character image data. This operation ensures that each character is processed independently, simplifies the normalization step, and improves small size text recognition results.

[0093] In a specific example, the corrected sequence data contains continuous text "Model C-100", the system is segmented into independent character images ("C", "-", "1", "0", "0"), and character image data is generated to facilitate subsequent standardization.

[0094] Step 3.5, process character image data based on normalization operation, unify size and direction, generate standardized character data to determine original text data.

[0095] Adjust the character image to a uniform size and direction, output standardized character data, and finally determine the original text data. "Normalization" refers to aligning size and direction; "standardized data" refers to consistent format input.

[0096] Process character image data (from step 3.4), apply scaling and rotation algorithms (such as bilinear interpolation), and unify all characters to a preset size (such as 32x32 pixels) to generate standardized character data. This step ensures that the OCR output original text data is accurate and supports structured parsing.

[0097] In a specific example, the size of the character image is not uniform, the system is uniformly scaled to the standard size, and standardized character data is generated, finally outputting the original text "Voltage 220V".

[0098] Step 4, structure the original text data to extract key device parameters and generate structured data; structured data includes voltage level, power parameters and manufacturer information.

[0099] Step 4.1: Process the raw text data based on a keyword positioning algorithm, identify predefined field keywords, and generate keyword position data.

[0100] Search for predefined keywords (e.g., "voltage" "power") in the text and mark their position information (e.g., starting index and length).

[0101] Predefine a keyword library (e.g., common fields for power equipment) and use a string matching algorithm (e.g., Trie tree) to quickly locate keywords and generate position data. This step should address text formatting issues and accurately identify parameter categories.

[0102] In a specific example, the original text "Model C-100 Voltage: 380V", the system locates the keyword "voltage" (position index 8-10) and generates position data, providing an anchor point for value extraction.

[0103] Step 4.2: Process the raw text data based on a regular expression matching algorithm, extract parameter values, and generate parameter value data.

[0104] Use regular expressions (rules that define text patterns) to extract values from the text (e.g., number + unit combination "380V").

[0105] Predefine regular rules (e.g., \d+[kK]?[Ww] to match power values) and combine them with keyword position data to accurately extract parameter values and generate standardized parameter value data (e.g., with a uniform unit "kW").

[0106] In a specific example, for "Power 50 kilowatts", the system applies regular rules to extract "50" and converts it to "50kW" to generate parameter value data.

[0107] Step 4.3: Process keyword position data and parameter value data based on a rule engine, associate fields with values, and generate preliminary structured data.

[0108] Pair keywords (field names) with extracted values through rule logic to form key-value pair data.

[0109] Predefine association rules (e.g., the keyword "voltage" followed by a value is the voltage level) to process keyword position and parameter value data, generate preliminary structured data (e.g., in dictionary format), and ensure data logical consistency.

[0110] In a specific example, the keyword "voltage" has a position index of 8-10, and the parameter value "380V" is located after it. The rule engine associates to generate {"voltage": "380V"}.

[0111] Step 4.4. Process the preliminary structured data based on template mapping operations, fill in the predefined database template, and generate structured data.

[0112] Map the preliminary data to a database standard template (e.g., SQL table structure) and output structured data that can be directly stored.

[0113] Call the device information template (e.g., HVAC device field template), map field names and value types (e.g., power to float), and generate structured data.

[0114] In a specific example, the preliminary data {"manufacturer": "B", "power": "50"} fills in the database template, generating structured records: {manufacturer: "B", power_kw: 50.0}.

[0115] Step 5. Calculate the energy consumption level of the device based on structured data, and combine with preset historical operation data and industry energy efficiency standards for energy efficiency evaluation, generating energy consumption evaluation results.

[0116] Step 5.1. Process structured data based on energy consumption modeling algorithms to construct device energy consumption models, generating model parameter data.

[0117] Use algorithms (e.g., linear regression or physical models) to construct prediction models based on device parameters (e.g., power, voltage), outputting model parameter data (e.g., energy consumption coefficient).

[0118] Process structured data (e.g., power parameters) based on preset algorithms to generate parameter data for subsequent evaluation. This step ensures objective energy consumption prediction and addresses the issue of device model diversity.

[0119] In a specific example, the system processes the structured data of the HVAC device {power: 50kW}, applies a linear regression model to generate parameter data (energy consumption coefficient per unit time is 2.4 kWh / kW), and predicts daily energy consumption to be 120 kWh.

[0120] Step 5.2. Process database historical operation data based on historical data query operations to obtain device operation efficiency indicators, generating historical efficiency data.

[0121] Retrieve device past operation records (e.g., monthly energy consumption) from a preset database, output historical efficiency data (e.g., average energy consumption or efficiency indicators). "Historical data" refers to stored old records; "efficiency indicators" refer to energy consumption and output ratio.

[0122] Query the database (e.g., SQL system) to obtain indicators such as load rate or energy consumption trend, generating historical efficiency data. This step provides a benchmark reference to improve evaluation accuracy.

[0123] In a specific example, the system queries 6 months of historical operation data of Device A, obtains an average energy consumption of 110 kWh, and generates historical efficiency data (efficiency rating "Medium") for subsequent comparison.

[0124] Step 5.3, Process model parameter data and historical efficiency data based on energy efficiency standard comparison algorithm, compare with industry threshold, generate energy efficiency gap data.

[0125] Compare model prediction data with historical data against industry standard threshold (e.g. energy efficiency rating), output energy efficiency gap data (e.g. percentage difference). Where "industry threshold" refers to standard value (e.g. ISO 50001); "gap data" refers to the degree of quantification of deviation.

[0126] Based on the difference value comparison data, generate gap data (e.g. "+15%"), identify potential energy waste points.

[0127] In a specific example, model parameter data predicts energy consumption of 120 kWh, industry standard threshold is 100 kWh, system generates energy efficiency gap data "+20%" after comparison, indicating high energy consumption.

[0128] Step 5.4, Process energy efficiency gap data based on multi-dimensional analysis operation, integrate environmental factors and load conditions, generate energy consumption evaluation results.

[0129] Integrate additional variables (e.g. ambient temperature or device load) and analyze gap data, output final evaluation results (e.g. "high energy consumption" rating). Where "multi-dimensional" refers to multiple influencing factors; "evaluation results" refers to comprehensive conclusions.

[0130] Based on a pre-set weight allocation strategy (e.g. temperature accounts for 30%, load accounts for 40%), process gap data, generate actionable evaluation results. This step ensures comprehensive evaluation and supports customized energy saving recommendations.

[0131] In a specific example, energy efficiency gap data +20%, combined with high summer temperature (environmental factor) and load rate 80%, system analysis generates evaluation results "current energy consumption level is high, suggest optimizing operation strategy".

[0132] Step 6, Generate energy saving optimization recommendation report based on energy consumption evaluation results; where the report includes device replacement, operation strategy adjustment or energy saving model recommendation.

[0133] Based on energy consumption evaluation results, automatically create customized optimization recommendation documents. Where "energy saving optimization" refers to solutions to reduce energy consumption; "report" refers to structured output documents containing text analysis, chart visualization and executable recommendations.

[0134] Through four steps of expert knowledge base matching, effectiveness prediction, report integration, and automatic verification (sub-steps 6.1-6.4), a fully automated closed loop from data to decision is realized. The report content strictly follows industry standards (such as building energy audit standards) to ensure that the recommendations are operable and support the intelligent transformation of building operation.

[0135] Step 6.1: Query the energy consumption evaluation results based on the expert knowledge base operation process, match the energy-saving scheme template, and generate initial suggestion data.

[0136] Query the pre-set expert knowledge base (including typical energy-saving cases), and match the most suitable solution template according to the evaluation results to output initial suggestion items.

[0137] Based on the pre-set knowledge base (such as energy-saving scheme library classified by equipment type), process the energy consumption evaluation results (such as "high energy consumption" rating), match the template (such as "transformer replacement scheme"), and generate initial suggestion data. This step ensures that the recommendations conform to engineering practice and avoid theoretical solutions that are detached from reality.

[0138] In a specific case, the evaluation result is "over-standard energy consumption of power distribution cabinet", the system matches the "old equipment replacement" template in the knowledge base, and generates the initial suggestion: "replace high-efficiency transformer (model Eco-T4000), estimated energy-saving rate 18%".

[0139] Step 6.2: Process the initial suggestion data and historical optimization records based on machine learning algorithms to predict the effectiveness of the suggestion and generate optimization probability data.

[0140] Apply machine learning models (such as random forest) to analyze historical implementation records and predict the success probability of the current suggestion, outputting quantitative probability data.

[0141] Train the model to analyze historical data (such as 10,000 optimization records), input the initial suggestion and context (such as building type, load rate), and generate probability data (such as "probability of energy-saving rate ≥15% is 82%"). This step improves the reliability of the suggestion through data-driven methods and reduces ineffective solutions.

[0142] In a specific case, the initial suggestion is "adjust the air conditioning operation strategy", and the model combines the historical data of B mall (similar scale building success cases) to generate optimization probability data "probability of saving 120,000 degrees per year is 85%".

[0143] Step 6.3: Process optimization probability data based on report generation engine, integrate text and charts, and generate report draft data.

[0144] Convert data into standard report format through the engine, including text description, data tables, and visual charts (such as energy consumption trend chart).

[0145] Call the report engine, structure and integrate three types of content: Assessment result summary (text); Optimization scheme comparison table (table); Energy saving benefit prediction graph (bar chart / line chart).

[0146] Generate a draft document in accordance with the ISO standard.

[0147] In a specific example, the engine converts "replace high-efficiency motor (probability 80%)" into a report chapter: text description of scheme details, table comparison of new and old equipment parameters, and chart showing 5-year energy saving benefit prediction.

[0148] Step 6.4, process report draft data based on automatic verification operation, verify data consistency, and generate energy saving optimization suggestions.

[0149] Check the consistency of parameters in the report (such as the matching degree of equipment power and energy consumption model), remove conflicting content, and output the final executable suggestions.

[0150] Pre-set verification rules: Uniform numerical units (such as all converted to kWh); Suggestions compatible with equipment model; Energy saving probability > threshold (such as 70%) to be retained.

[0151] In a specific example, the power parameter of the "lighting renovation scheme" in the draft report conflicts with the original data, and the system automatically corrects and generates the final suggestion: "replace LED lamps (model L-200), match actual power 15kW, expected annual power saving 24,000 degrees".

[0152] The above is the method embodiment of the present application. Based on the same inventive concept, the building operation and maintenance equipment based on OCR recognition and energy efficiency analysis is also provided by the embodiments of the present application, and the structure is as shown in Figure 2 .

[0153] Figure 2 The building operation and maintenance equipment based on OCR recognition and energy efficiency analysis provided by the embodiments of the present application is shown in Figure 2 , which includes: At least one processor 201; And the memory 202 in communication connection with the at least one processor; Among them, the memory 202 stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to: An image acquisition device is used to acquire a power equipment nameplate image; the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; the image is preprocessed to perform grayscale, binarization, contrast enhancement, and blur removal, and preprocessed image data is generated; the preprocessed image data is processed based on a preset deep learning OCR model to identify the text content in the image, and original text data is generated; the original text data is structurally parsed to extract equipment key parameters, and structured data is generated; the structured data includes voltage level, power parameters, and manufacturer information; the equipment energy consumption level is calculated based on the structured data, and energy efficiency is evaluated in combination with preset historical operation data and industry energy efficiency standards, and an energy consumption evaluation result is generated; an energy saving optimization suggestion report is generated according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment, or energy saving model recommendation.

[0154] Some embodiments of the present application provide a non-volatile computer storage medium for building operation and maintenance based on OCR recognition and energy efficiency analysis corresponding to Figure 1 The computer executable instructions are set as follows: An image acquisition device is used to acquire a power equipment nameplate image; the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; the image is preprocessed to perform grayscale, binarization, contrast enhancement, and blur removal, and preprocessed image data is generated; the preprocessed image data is processed based on a preset deep learning OCR model to identify the text content in the image, and original text data is generated; the original text data is structurally parsed to extract equipment key parameters, and structured data is generated; the structured data includes voltage level, power parameters, and manufacturer information; the equipment energy consumption level is calculated based on the structured data, and energy efficiency is evaluated in combination with preset historical operation data and industry energy efficiency standards, and an energy consumption evaluation result is generated; an energy saving optimization suggestion report is generated according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment, or energy saving model recommendation.

[0155] Each of the embodiments in the present application is described in a progressive manner, and the same or similar parts of each of the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the Internet of Things device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiment.

[0156] The system and medium provided by the embodiments of the present application are one-to-one corresponding, and therefore the system and medium also have similar beneficial technical effects to the method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the system and medium will not be described here again.

[0157] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. In addition, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0158] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.

[0159] These computer program instructions can also be stored in a computer-readable memory capable of causing a computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.

[0160] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The means for implementing the functions specified in one or more flows and / or blocks.

[0161] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memories.

[0162] Memory can include non-persistent memory, Random Access Memory (RAM), and / or non-volatile memory, such as Read Only Memory (ROM) or flash memory, in computer readable media. Memory is an example of computer readable media.

[0163] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0164] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not include only those elements in the list, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0165] The above only is an embodiment of the present application, and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A building operation and maintenance method based on OCR identification and energy efficiency analysis, characterized in that, The method comprises: acquiring a power equipment nameplate image through an image acquisition device; wherein the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image comprises a nameplate text area and equipment identification information; preprocessing the power equipment nameplate image to perform grayscale, binarization, contrast enhancement and blur removal, and generating preprocessed image data; processing the preprocessed image data based on a preset deep learning OCR model to identify the text content in the image and generate original text data; structurally analyzing the original text data to extract equipment key parameters and generate structured data; wherein the structured data comprises voltage level, power parameter and manufacturer information; calculating the equipment energy consumption level based on the structured data, and combining the preset historical operation data and industry energy efficiency standard to perform energy efficiency evaluation and generate energy consumption evaluation results; generating an energy saving optimization suggestion report according to the energy consumption evaluation results; wherein the report comprises equipment replacement, operation strategy adjustment or energy saving model recommendation.

2. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 1, characterized in that, The preprocessing of the power equipment nameplate image to perform grayscale, binarization, contrast enhancement and blur removal, and generating preprocessed image data specifically comprises: processing the power equipment nameplate image based on a preset grayscale conversion algorithm to convert the color image into grayscale image data; processing the grayscale image data based on a preset histogram equalization algorithm to enhance the image contrast and generate enhanced image data; processing the enhanced image data based on a preset adaptive threshold segmentation algorithm to perform binarization operation and generate binary image data; processing the binary image data based on a preset Gaussian filtering algorithm to remove image noise and blur and generate clear image data; correcting the inclination angle of the clear image data to generate preprocessed image data.

3. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 2, characterized in that, The processing of the grayscale image data based on a preset histogram equalization algorithm to enhance the image contrast and generate enhanced image data specifically comprises: processing the grayscale image data based on global histogram equalization to adjust the overall brightness distribution and generate equalized image data; processing the equalized image data based on adaptive histogram equalization to enhance the local area in zones and generate adaptive image data; processing the adaptive image data based on gamma correction algorithm to optimize the brightness curve and generate corrected image data; processing the corrected image data based on sharpening filtering to enhance the text edge and generate enhanced image data.

4. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 2, characterized in that, The processing of the binary image data based on a preset Gaussian filtering algorithm to remove image noise and blur and generate clear image data specifically comprises: processing the binary image data based on Gaussian kernel construction operation to define the filter window size and generate kernel parameter data; processing the binary image data and kernel parameter data based on convolution operation to smooth the pixel value and generate smoothed image data; processing the smoothed image data based on edge detection algorithm to identify the text contour and generate edge map data; processing the edge map data based on noise suppression operation to remove isolated noise points and generate clear image data.

5. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 1, characterized in that, processing the preprocessed image data based on a preset deep learning OCR model to identify text content in the image and generate raw text data, specifically including: processing the preprocessed image data based on a convolutional neural network to extract character feature vectors and generate feature map data; processing the feature map data based on a recurrent neural network to model text sequence dependency and generate sequence prediction data; processing the sequence prediction data based on an attention mechanism to focus on key text areas and generate corrected sequence data; processing the corrected sequence data based on a character segmentation algorithm to separate independent character images and generate character image data; processing the character image data based on a normalization operation to unify size and direction and generate standardized character data to determine the raw text data.

6. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 1, characterized in that, performing structured analysis on the raw text data to extract device key parameters and generate structured data, specifically including: processing the raw text data based on a keyword positioning algorithm to identify pre-defined field keywords and generate keyword position data; processing the raw text data based on a regular expression matching algorithm to extract parameter values and generate parameter value data; processing the keyword position data and parameter value data based on a rule engine to associate fields and values and generate preliminary structured data; processing the preliminary structured data based on a template mapping operation to fill in a pre-defined database template and generate structured data.

7. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 1, characterized in that, calculating device energy consumption levels based on the structured data, and combining preset historical operation data and industry energy efficiency standards to perform energy efficiency evaluation and generate energy consumption evaluation results, specifically including: processing the structured data based on an energy consumption modeling algorithm to construct a device energy consumption model and generate model parameter data; processing database historical operation data based on a historical data query operation to obtain device operation efficiency indicators and generate historical efficiency data; processing the model parameter data and historical efficiency data based on an energy efficiency standard comparison algorithm to compare with industry thresholds and generate energy efficiency gap data; processing the energy efficiency gap data based on a multi-dimensional analysis operation to generate energy consumption evaluation results by comprehensively considering environmental factors and load conditions.

8. The building operation and maintenance method based on OCR identification and energy efficiency analysis according to claim 1, characterized in that, generating an energy-saving optimization recommendation report based on the energy consumption evaluation results; wherein the report includes device replacement, operation strategy adjustment, or energy-saving model recommendation, specifically including: processing the energy consumption evaluation results based on an expert knowledge base query operation to match energy-saving scheme templates and generate initial suggestion data; processing the initial suggestion data and historical optimization records based on a machine learning algorithm to predict suggestion effectiveness and generate optimization probability data; processing the optimization probability data based on a report generation engine to integrate text and charts and generate report draft data; processing the report draft data based on an automatic verification operation to verify data consistency and generate energy-saving optimization recommendations.

9. A building operation and maintenance device based on OCR identification and energy efficiency analysis, characterized in that, The device includes: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: An image acquisition device is used to acquire a power equipment nameplate image; the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; The image is preprocessed to perform grayscale, binarization, contrast enhancement, and blur removal, to generate preprocessed image data; The preprocessed image data is processed based on a preset deep learning OCR model to identify the text content in the image, to generate original text data; The original text data is structurally parsed to extract equipment key parameters, to generate structured data; the structured data includes voltage level, power parameters, and manufacturer information; The equipment energy consumption level is calculated based on the structured data, and energy efficiency is evaluated in combination with preset historical operation data and industry energy efficiency standards, to generate an energy consumption evaluation result; An energy-saving optimization suggestion report is generated according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment, or energy-saving model recommendation. 10.A non-transitory computer storage medium storing computer-executable instructions for building operation and maintenance based on OCR identification and energy efficiency analysis, the computer-executable instructions comprising instructions for: The computer executable instructions are configured to: ​ An image acquisition device is used to acquire a power equipment nameplate image; the power equipment nameplate image is captured by the image acquisition device under natural light or adverse weather conditions, and the power equipment nameplate image includes a nameplate text area and equipment identification information; The image is preprocessed to perform grayscale, binarization, contrast enhancement, and blur removal, to generate preprocessed image data; The preprocessed image data is processed based on a preset deep learning OCR model to identify the text content in the image, to generate original text data; The original text data is structurally parsed to extract equipment key parameters, to generate structured data; the structured data includes voltage level, power parameters, and manufacturer information; The equipment energy consumption level is calculated based on the structured data, and energy efficiency is evaluated in combination with preset historical operation data and industry energy efficiency standards, to generate an energy consumption evaluation result; An energy-saving optimization suggestion report is generated according to the energy consumption evaluation result; the report includes equipment replacement, operation strategy adjustment, or energy-saving model recommendation.

Images