Substation engineering method based on importing Excel and complementing 3D models

Abstract

The present invention provides a substation engineering method based on importing Excel and complementing with a three-dimensional model, comprising the following steps: S1 divides the project into multiple branch projects according to a preliminary design plan; S2 collects historical project completion schedules, material attributes and costs, constructs material attribute tables and cost tables, and constructs data on time, material attributes, and costs; S3 completes the mapping of data to each sub-project, block project, and detail project, and completes the visualization process simulation of engineering construction for multiple separated branches; S4 then constructs a bimodal deep learning artificial intelligence model for modifying the preliminary design plan. The present invention realizes the simulation of the visualization engineering process, improves the intuitive analyzability of the engineering progress, and provides a visualized and convenient reference for the optimization of engineering construction design and the comprehensive evaluation of the engineering.

Description

Technical Field

[0001] The invention relates to a substation engineering method, in particular to a substation engineering method based on importing Excel and complementing a three-dimensional model, and belongs to the field of engineering intelligence. Background Art

[0002] The construction of a project requires a certain construction sequence, and each type of construction plan has different material support and cost accounting. The existing engineering method maintains material attributes through Excel and further realizes attribute recognition through the so-called ladder table, and calculates the cost based on this. In essence, it does not use the three-dimensional model for design. It only uses the table data to perform the function mapping process of attribute-cost, and cannot simulate its construction progress.

[0003] The 3D model is a means of displaying the project in space. However, the logical distinction of the construction plan cannot vividly reflect the construction progress and the current completion cost, so 3D display becomes a possible means to solve this problem. However, how to visualize the project progress and cost, how to predict multi-index project evaluation through cost and progress, and even make detailed design summary and improvement have become problems that need to be solved by existing technologies. Summary of the invention

[0004] Based on the defects of the above-mentioned prior art, the present invention starts with the construction of a three-dimensional model, integrates material attribute cost data, and project progress visualization into the three-dimensional model, so as to simulate the project progress and cost. In view of the above-mentioned overall design scheme, the present invention provides a substation engineering method based on importing Excel and complementing the three-dimensional model, including the following steps:

[0005] S1: Divide the project into a plurality of sub-projects according to the preliminary design plan, and divide at least one sub-project into a plurality of block projects, and further divide at least one block project into a plurality of detail projects, or only divide the project into a plurality of sub-projects, or divide the project into a plurality of sub-projects and at least one sub-project into at least one block project or at least one detail project;

[0006] S2 collects historical project completion schedules, material attributes and costs, and constructs a material attribute table and a cost table, wherein the material attribute table contains material physical and chemical attribute classifications and product supplier classifications, and the cost table contains real-time market prices of each material product under each supplier classification corresponding to each physical and chemical attribute classification, which can be an Excel table;

[0007] S3 maps the timetable, material attribute table, and cost table to each sub-project, block project, and detail project, i.e., three projects, and performs construction simulation from the beginning until the three-dimensional model is completed. According to the difference of each material product, multiple three-project construction simulation visualization processes are separated. When the simulation of each detail project, block project, and sub-project is completed, the current cost of each corresponding three projects is recorded;

[0008] S4 builds a bimodal deep learning artificial intelligence model, using different schedules, material product costs, the generated text data, and the 3D graphic data of the three-dimensional model to learn the best plan for modifying the preliminary design plan.

[0009] Among them, optionally, the timetable, material attribute table, and cost table are mapped to each sub-project, block project, and detail project, specifically including marking the time data in the timetable, material attribute table, and cost table, the attribute data of the physical and chemical property classification, the product supplier classification, and the cost data of the real-time market price into the three-dimensional model of each of the three projects in the current process, and displaying them by clicking the mark.

[0010] Optionally, the viewing angle of the three-dimensional model can be changed, and the zooming in and out method can be used to display the visual process.

[0011] Optionally, the time data in the historical project completion schedule includes the historical construction completion time required for the three projects, and the construction sequence arrangement of the three projects (or multiple sub-projects when the project is only divided into multiple sub-projects).

[0012] Optionally, the engineering construction simulation is performed from the beginning until the three-dimensional model is constructed, specifically including confirming the three-dimensional model components and three-dimensional model equipment corresponding to the materials required in the three projects, and completing the construction of each component and equipment according to the time data.

[0013] Optionally, the construction includes the building, assembly and equipment deployment of components.

[0014] Optionally, recording the current cost of each of the three corresponding projects includes expanding and recording real-time cost information at the mark, or highlighting part of the cost data that has been completed in the cost data.

[0015] Optionally, the S4 specifically includes:

[0016] S4-1 converts the time data and the cost data into first text data and second text data respectively, and converts the three-dimensional model components and three-dimensional model equipment corresponding to the components and equipment of the three projects in the three-dimensional model into first 3D graphic data and second 3D graphic data respectively;

[0017] S4-2 constructs the first text data and the second text data for respectively training the long short-term memory models LSTM1 and LSTM2, and the first 3D graphics data and the second 3D graphics data for respectively training the first 3D convolutional neural networks 3DCNN1 and 3DCNN2, and respectively forms the first training set to the fourth training set, and the corresponding first verification set to the fourth verification set with the first text data, the second text data, the first 3D graphics data and the second 3D graphics data, and the corresponding ratio of the training set to the verification set is 7-4:1-3, and the corresponding engineering multi-index comprehensive evaluation level of the training set and the verification set is used as a label; wherein the first 3D graphics data and the second 3D graphics data are formed by three-dimensional construction and texture rendering processing of different material products with different physical images, so as to be used as information data for the training.

[0018] S4-3 inputs the weighted outputs of LSTM1, LSTM2, 3DCNN1, and 3DCNN2 into the CONACT function, and finally outputs the classification of the engineering multi-index comprehensive evaluation level using the softmax function, so as to use the LSTM1, LSTM2, 3DCNN1, and 3DCNN2 inputs corresponding to each training set input for training, and use the corresponding verification set to perform a comprehensive evaluation of the level accuracy, so as to continuously optimize each network parameter until the accuracy stabilizes and the training is stopped, thus obtaining a trained dual-modal deep learning artificial intelligence model.

[0019] Therefore, the fusion learning and classification of multi-class network models using different data texts and 3D graphics are used to complete the prediction of the completion quality of the project. The comprehensive evaluation includes statistical evaluation of multiple indicators such as the duration of the project, whether it is non-standard construction (for example, the use of over-standard, sub-standard or non-famous brand materials and equipment), acceptance report data, failure rate during operation, user complaints, etc. It can be seen that only using dual-modal data training can predict the prediction of comprehensive evaluation of more than two indicators, thus saving the cost of evaluation. However, there is an upper limit to the indicator, and more indicators cannot accurately predict the evaluation level, which is very different from the actual evaluation level and loses the meaning of prediction.

[0020] Therefore, preferably, the number of multiple indicators is 3-5.

[0021] Preferably, the separation is to select material specifications and corresponding cost ranges according to the standards of the project, and materials and materials products corresponding to the specifications and costs that do not meet the standards are not separated. That is, each material in the multiple three-project construction simulation visualization processes separated meets the standards, and the corresponding process is not simulated for those that do not meet the standards. For example, the cement grade in the concrete used in the construction of the substation room cannot be the cement grade for building a super high-rise building, so the simulation of the room completed with the cement grade of the super high-rise building is not separated and displayed.

[0022] Preferably, only the first text data or the second text data, and the corresponding first 3D graphics data or the second 3D graphics data are selected for the training and verification.

[0023] Optionally, the modification of the preliminary design plan includes modifying the division of the three projects, modifying materials and cost standards, and modifying time data.

[0024] Optionally, continuously accumulate new project completion schedules, material attributes and costs, update the training set, update the multi-index comprehensive evaluation grade classification, and continuously update the training of LSTM1, LSTM2, 3DCNN1, and 3DCNN2. Beneficial Effects

[0025] On the one hand, through the subdivision of the project, 3D model visualization, as well as project construction time data, material attribute data, and cost data, a visual simulation of the project process is achieved, and the current cost can be viewed in real time.

[0026] On the other hand, it uses time data, material data, and cost data as text data, and uses the 3D model of the material based on different material products as 3D image data. The two data modalities are used to build a dual-modal deep learning artificial intelligence model to predict the final comprehensive evaluation of the project based on more indicators under the current design of materials, costs, and construction periods. Overall, it improves the intuitive analyzability of the project progress and provides a visual and convenient reference for the optimization of project construction design and comprehensive project evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 a 3D model diagram from a certain perspective when the project is completed, Figure 1 b- Figure 1 c are the 3D model schematic diagrams of the building block project and equipment block project of a certain separated branch when completed, Figure 1 d is Figure 1 b and Figure 1 Schematic diagram of the 3D model corresponding to the visualization process of the 2-branch and 4-branch separation engineering in c.

[0028] Figure 2 Construct a dual-modal deep learning artificial intelligence model flow chart, where:

[0029] 1-4 are sub-project numbers, AE are the first station area control center, the second station area control center, the first station area computer room, the second station area computer room, and the extended equipment area computer room, respectively; FH are the first station area equipment, the second station area equipment, and the extended equipment area equipment, respectively. DETAILED DESCRIPTION

[0030] Based on Figure 1 Taking a 3D model schematic diagram of a substation as an example, the intelligent method for visualizing substation engineering includes the following steps:

[0031] S1 is a preliminary plan, which is divided into sub-project 1, the first station area, sub-project 2, the second station area, sub-project 3, the extended equipment area, and sub-project 4, the motor vehicle lane in the station. The control center A and the computer room C of the first station area, the control center B and the computer room D of the second station area, and the computer room E of the extended equipment area are respectively regarded as building block projects. The equipment F of the first station area, the equipment G of the first station area, and the equipment H of the third station area are regarded as equipment block projects. And the monitoring equipment in the computer room C-computer room D, the detection and maintenance instruments and equipment ( Figure 1 (not shown) configuration requirements and placement methods, specific installation methods of equipment F-equipment H, such as the provision of auxiliary tooling and auxiliary facilities, etc. as detail projects. Figure 1 The boundaries of the building and equipment blocks are divided by red dotted lines for display and distinction in the simulation project progress.

[0032] S2 collects the historical project completion schedules of other substations, as well as material attributes and costs. According to national planning, more than 1,000 smart substations will be added each year. Therefore, sufficient small sample training data can be obtained in advance. As time goes by, the number of samples will be expanded to update the dual-modal deep learning artificial intelligence model in the following text.

[0033] In this step, it is necessary to construct a material attribute table and a cost table. The material attribute table contains the physical and chemical attribute classification of the material and the product supplier classification. The cost table contains the real-time market price of each material product under each supplier classification corresponding to each physical and chemical attribute classification.

[0034] For example, the material products provided by each supplier have certain physical and chemical properties. For those used in building, they include bricks, cement, sand, steel bars, and office water, electricity and gas line control; for the construction of substation equipment, they include various electrical components, wires, testing and maintenance instruments and equipment, and monitoring equipment, as well as on-site safety monitoring devices.

[0035] Among them, the time required for the completion of the three projects and the construction sequence of the three projects are used as the content of the time data. In this embodiment, the sequence is as follows: Figure 1 a- Figure 1 As shown in c, complete first Figure 1 b. The building block works and the motorway within the station, and then complete Figure 1 c's sub-project 1 and the device block project for device F and device G in sub-project 2, and finally complete the project for extended device H in sub-project 3, as shown in the following figure: Figure 1 Complete the entire project as shown in a. Figure 1 In the project progress of b, the equipment operation and debugging work can be started in the early stage, and the second phase is to build the substation equipment and power line control, and start the formal substation service trial operation. The third phase will carry out extended functional services, such as customized services with special requirements for power parameters.

[0036] S3 maps the timetable, material attribute table, and cost table to the above S1 step. Figure 1 Each sub-project 1 to sub-project 4, each block project, and detail project shown in the figure shall be simulated from the beginning until the 3D model is constructed, specifically including first confirming the materials needed to construct the control center, computer room, equipment, monitoring equipment layout, and detection and maintenance instruments and equipment in S1 in the three projects, as well as the 3D model components and 3D model equipment corresponding to these materials constructed through real images, and then completing the construction of each component and equipment according to the time data.

[0037] When the simulation of the three projects is completed, the current cost of each of the three projects is recorded, and according to the difference of each material product, multiple visualization processes of the three project construction simulation are separated. Figure 1 When constructing according to the requirements of power capacity, anti-interference requirements, substation size, etc., multiple separate synchronous visualization processes can be adopted within the predetermined fluctuation range of the established standards.

[0038] like Figure 1 b shows an example of a 3D model diagram of a separation branch when the building block project and the motorway sub-project 4 are completed. Figure 1 c is an example of a 3D model diagram of a separate branch when sub-project 1, sub-project 2, and sub-project equipment block project are completed. Figure 1 d gives Figure 1 b and Figure 1 c are examples of 3D model diagrams with two and four separation branches. Figure 1 The 2×3 array of separated branches of d can simultaneously visualize the engineering progress simulation monitoring, and can analyze in real time whether the construction period is delayed and the current cost.

[0039] Specifically, the timetable, material attribute table, and cost table are mapped to each sub-project, block project, and detail project, including marking the time data in the timetable, material attribute table, and cost table, the physical and chemical attribute classification, the attribute data of the product supplier classification, and the cost data of the real-time market price to the 3D model of each of the three projects in the current process, and clicking in real time Figure 1 b- Figure 1 The letter mark AH in d can pop up a dialog box to select data type, search, and view the time data, attribute data, and cost data for display ( Figure 1 not shown).

[0040] S4 builds a bimodal deep learning artificial intelligence model, which will form text data based on different progress schedules, material product costs, and 3D graphic data of the three-dimensional model to learn the best plan for modifying the preliminary design plan.

[0041] Specific as Figure 2 As shown, S4-1 converts the time data and the cost data into first text data and second text data respectively, and converts the three-dimensional model components and three-dimensional model equipment corresponding to each component and equipment of the three projects in the three-dimensional model into first 3D graphic data and second 3D graphic data respectively;

[0042] S4-2 constructs the first text data and the second text data for respectively training the long short-term memory models LSTM1 and LSTM2, and the first 3D graphics data and the second 3D graphics data for respectively training the first 3D convolutional neural networks 3DCNN1 and 3DCNN2, and respectively forms the first training set to the fourth training set, and the corresponding first verification set to the fourth verification set with the first text data, the second text data, the first 3D graphics data and the second 3D graphics data, and the corresponding ratio of the training set to the verification set is 7-4:1-3, and the corresponding engineering multi-index comprehensive evaluation level of the training set and the verification set is used as a label; wherein the first 3D graphics data and the second 3D graphics data are formed by three-dimensional construction and texture rendering processing of different material products with different physical images, so as to be used as information data for the training.

[0043] S4-3 weights the outputs of LSTM1, LSTM2, 3DCNN1, and 3DCNN2 by q1, q2, q3, and q4, respectively, and inputs them into the CONACT function (brown frame). Finally, the output uses the softmax function (light green frame) to classify the engineering multi-index comprehensive evaluation level, so as to use each training set to input LSTM1, LSTM2 (purple frame), and 3DCNN1, 3DCNN2 (blue frame) input for training, and use the corresponding verification set to perform comprehensive evaluation level accuracy, so as to continuously optimize each network parameter until the accuracy stabilizes and stops training, and obtains a trained dual-modal deep learning artificial intelligence model. This embodiment uses text data and 3D graphic data as two data modalities ( Figure 2 In the figure, the network models long short-term memory and convolutional neural network suitable for text and graphics were used to predict the evaluation of multiple indicators.

[0044] By continuously accumulating new project completion schedules, for example, the construction block project and the equipment block project can be carried out at the same time, so that the completion time data of the two block projects will be closer. In addition, it is necessary to update material attributes and their costs, update the training set, and update the multi-index comprehensive evaluation level classification, and continuously update the training of LSTM1, LSTM2, 3DCNN1, and 3DCNN2.

Claims

1. A substation engineering method based on importing Excel and complementing the 3D model, characterized in that: The steps include: S1: Divide the project into a plurality of sub-projects according to the preliminary design plan, and divide at least one sub-project into a plurality of block projects, and further divide at least one block project into a plurality of detail projects, or only divide the project into a plurality of sub-projects, or divide the project into a plurality of sub-projects and at least one sub-project into at least one block project or at least one detail project; S2 collects historical project completion schedules, material attributes and cost data, and constructs a material attribute table and a cost table, wherein the material attribute table contains material physical and chemical attribute classifications and product supplier classifications, and the cost table contains real-time market prices of each material product under each supplier classification corresponding to each physical and chemical attribute classification; the historical project completion schedule contains time data, and the time data includes the historical construction completion time required for the three projects, and the construction sequence arrangement of the three projects, or when the project is only divided into multiple sub-projects, the construction sequence arrangement of multiple sub-projects; S3 maps the timetable, material attribute table, and cost table to each sub-project, block project, and detail project, i.e., three projects, and performs construction simulation from the beginning until the three-dimensional model is completed. According to the difference of each material product, multiple three-project construction simulation visualization processes are separated, and when the simulation of each detail project, block project, and sub-project is completed, the current cost of each corresponding three projects is recorded; S4 builds a bimodal deep learning artificial intelligence model, and uses the text data formed by different schedules and material product costs and the 3D graphic data of the three-dimensional model to learn the best plan for modifying the preliminary design plan, wherein the modification of the preliminary design plan includes modifying the division of the three projects, modifying the material and cost standards, and modifying the time data; The S4 specifically includes: S4-1 converts the time data and the cost data into first text data and second text data respectively; converts the three-dimensional model components and three-dimensional model equipment corresponding to the components and equipment of the three projects in the three-dimensional model into first 3D graphic data and second 3D graphic data respectively; S4-2 constructs the first text data and the second text data for respectively training the long short-term memory models LSTM1 and LSTM2, and the first 3D graphic data and the second 3D graphic data for respectively training the first 3D convolutional neural network 3DCNN1 and the second 3D convolutional neural network 3DCNN2; respectively forms the first training set to the fourth training set corresponding to the first text data, the second text data, the first 3D graphic data and the second 3D graphic data, and the corresponding first verification set to the fourth verification set, with the corresponding engineering multi-index comprehensive evaluation level of the training set and the verification set as the label; wherein the first 3D graphic data and the second 3D graphic data are obtained by different material products through Different physical images are three-dimensionally constructed and textured to form information data for the training; S4-3 weightedly input the output ends of LSTM1, LSTM2, 3DCNN1, and 3DCNN2 into the CONACT function, and finally the output is classified by the softmax function for the comprehensive evaluation grade of the engineering multi-index, so as to use the LSTM1, LSTM2, 3DCNN1, and 3DCNN2 input ends corresponding to each training set input for training, and use the corresponding verification set to perform a comprehensive evaluation grade accuracy, so as to continuously optimize each network parameter until the accuracy stabilizes and the training is stopped, so as to obtain a trained dual-modal deep learning artificial intelligence model.

2. The method according to claim 1, characterized in that The construction simulation is carried out from the beginning until the 3D model is completed, including confirming the 3D model components and 3D model equipment corresponding to the materials required in the three projects, and completing the construction of each component and equipment according to the time data.

3. The method according to claim 2, characterized in that The construction includes the building, assembly and equipment installation of components.

4. The method according to claim 3, characterized in that Recording the current cost of each of the three corresponding projects includes expanding and recording the real-time cost information at the mark, or highlighting the completed part of the cost data in the cost data.

5. The method according to claim 1, characterized in that The separation is to select material specifications and corresponding cost ranges according to the engineering standards, and the material products corresponding to the materials and costs that do not meet the specifications of the standards will not be separated; the number of indicators of the multiple indicators is 3-5; by continuously accumulating new project completion schedules, material attributes and costs, the training set is updated, and the multi-indicator comprehensive evaluation grade classification is updated, and the training of LSTM1, LSTM2, 3DCNN1, and 3DCNN2 is continuously updated.

6. The method according to claim 1, characterized in that Only the first text data or the second text data, and the corresponding first 3D graphics data or the second 3D graphics data are selected for the training and verification.

Images