Plateau area construction risk early warning method, device and equipment and readable storage medium
By constructing a construction risk early warning model in plateau regions and using data from plain regions for pre-training and differential calculation, the problem of insufficient construction safety early warning data in plateau regions was solved, achieving rapid and accurate early warning results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-03-13
AI Technical Summary
During the construction of new energy power plants in plateau areas, due to limited data, existing technologies are unable to provide rapid and accurate early warnings of construction safety.
By acquiring a construction risk dataset from the plains region for pre-training, calculating the maximum mean difference between the source and target domain data, and employing transfer learning or fine-tuning methods based on the magnitude of the difference, a risk early warning model for the plateau region is constructed.
In situations where data is scarce, a highly accurate construction risk early warning model can be quickly built, adapted to the plateau environment, and its accuracy and adaptability can be improved.
Smart Images
Figure CN121660428A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of construction safety technology in plateau areas, and in particular to a method, device, equipment and readable storage medium for early warning of construction risks in plateau areas. Background Technology
[0002] Plateau regions (such as the Qinghai-Tibet Plateau) have unique geographical and climatic characteristics such as high altitude, low oxygen, low temperature and strong ultraviolet radiation, which poses a severe challenge to the construction safety of new energy power plants (wind farms and photovoltaic power plants). Timely and accurate construction safety early warning is essential.
[0003] The construction period for new energy power plants in plateau regions is usually only 3-6 months, resulting in limited construction data collection. Furthermore, historical data for construction safety early warning in plateau regions is extremely limited. Therefore, how to quickly and accurately provide early warnings for construction safety in plateau regions with limited data is an urgent problem to be solved. Summary of the Invention
[0004] This application provides a method, device, equipment, and readable storage medium for early warning of construction risks in plateau areas, aiming to solve the technical problem of how to quickly and accurately provide early warning of construction safety in plateau areas when data is limited.
[0005] In a first aspect, embodiments of this application provide a method for early warning of construction risks in plateau areas, the method comprising: We obtained a construction risk dataset of new energy power plants in the plains area as the source domain data, and used the source domain data to pre-train the random forest algorithm model to obtain a pre-trained model. Data on construction risks of new energy power plants in plateau areas were collected as the target domain data. Calculate the maximum mean difference between the data distributions of the source domain data and the target domain data; If the maximum mean difference is greater than the preset threshold, the target domain data is used to perform transfer learning on the pre-trained model to obtain a risk warning model for the plateau region. If the maximum mean difference is not greater than the preset threshold, the pre-trained model is fine-tuned using target domain data to obtain a risk warning model for plateau regions.
[0006] Optionally, the step of using target domain data to perform transfer learning on the pre-trained model to obtain a risk warning model for plateau areas includes: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
[0007] Optionally, the method for early warning of construction risks in plateau areas also includes: Calculate the new maximum mean difference between the data distributions of the source domain data and the target domain data at a preset period; The transfer learning weight λ is increased or decreased accordingly based on whether the new maximum mean difference increases or decreases.
[0008] Optionally, the larger the amount of data in the target domain, the smaller the corresponding transfer learning weight λ should be.
[0009] Optionally, the step of fine-tuning the pre-trained model using target domain data to obtain a risk warning model for plateau regions includes: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
[0010] Secondly, embodiments of this application provide a construction risk early warning device for high-altitude areas, the high-altitude area construction risk early warning device comprising: The acquisition module is used to acquire the construction risk dataset of new energy power plants in the plains area as the source domain data, and to use the source domain data to pre-train the random forest algorithm model to obtain the pre-trained model. The data acquisition module is used to collect construction risk datasets for new energy power plants in plateau areas as target domain data. The calculation module is used to calculate the maximum mean difference between the data distributions of the source domain data and the target domain data; The transfer module is used to perform transfer learning on the pre-trained model using target domain data if the maximum mean difference is greater than a preset threshold, so as to obtain a risk warning model for plateau areas. The fine-tuning module is used to fine-tune the pre-trained model using target domain data if the maximum mean difference is not greater than a preset threshold, thereby obtaining a risk warning model for plateau regions.
[0011] Optionally, the migration module is used for: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
[0012] Optionally, the fine-tuning module is used for: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
[0013] Thirdly, this application provides a high-altitude construction risk early warning device, which includes a processor, a memory, and a high-altitude construction risk early warning program stored in the memory and executable by the processor. When the high-altitude construction risk early warning program is executed by the processor, it implements the steps of the high-altitude construction risk early warning method described above.
[0014] Fourthly, embodiments of this application provide a readable storage medium storing a high-altitude construction risk warning program, wherein when the high-altitude construction risk warning program is executed by a processor, it implements the steps of the high-altitude construction risk warning method as described above.
[0015] The beneficial effects of the technical solutions provided in this application include: In this embodiment, a construction risk dataset of new energy power plants in plain areas is obtained as the source domain data, and the random forest algorithm model is pre-trained using the source domain data to obtain a pre-trained model; a construction risk dataset of new energy power plants in plateau areas is collected as the target domain data; the maximum mean difference between the data distributions of the source domain data and the target domain data is calculated; if the maximum mean difference is greater than a preset threshold, the pre-trained model is transferred to the target domain data to obtain a risk warning model for plateau areas; if the maximum mean difference is not greater than the preset threshold, the pre-trained model is fine-tuned using the target domain data to obtain a risk warning model for plateau areas. Through the embodiments of this application, when target domain data is scarce in plateau areas, a pre-trained model is obtained by pre-training the random forest algorithm model using source domain data from plains areas. Then, the difference in data distribution between the source domain data and the target domain data is compared. When the distribution difference is large, transfer learning is used to enhance the model's adaptability to the plateau environment. When the difference is small, fine-tuning is used to retain the pre-training advantages of the source domain. Thus, only a small amount of target domain data is needed to quickly and accurately obtain a risk warning model for plateau areas based on the pre-trained model. This solves the problem of scarce data in plateau areas and has the advantages of rapid deployment and high warning accuracy. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating an embodiment of the construction risk early warning method for plateau areas according to this application; Figure 2 This is a schematic diagram of the functional modules of an embodiment of the high-altitude construction risk early warning device of this application; Figure 3 This is a schematic diagram of the hardware structure of the high-altitude construction risk early warning device involved in the embodiments of this application. Detailed Implementation
[0017] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.
[0018] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0019] Firstly, embodiments of this application provide a method for early warning of construction risks in plateau areas.
[0020] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the construction risk early warning method for plateau areas according to this application. Figure 1 As shown, the methods for early warning of construction risks in plateau areas include: Step S10: Obtain the construction risk dataset of new energy power plants in the plain area as the source domain data, and use the source domain data to pre-train the random forest algorithm model to obtain the pre-trained model.
[0021] In this embodiment, a construction risk dataset of new energy power plants in plain areas (such as the North China Plain and the middle and lower reaches of the Yangtze River Plain) is first obtained from a historical database as the source domain data for transfer learning. This dataset contains various risk factors during construction (such as weather conditions, equipment failure rates, and personnel operational errors) and their corresponding risk levels. Due to the long history of new energy power plant construction in plain areas and the rich accumulation of data, this dataset typically contains over 100,000 records. Then, this dataset is used to pre-train a random forest algorithm model. By adjusting hyperparameters such as the number of trees and maximum depth, a basic model with strong generalization ability is constructed—the pre-trained model. Random forest, as an ensemble learning method, can effectively handle high-dimensional features and improve the robustness of the model through the ensemble of multiple trees. Through pre-training, the model has learned the general patterns and characteristics of construction risks, laying the foundation for subsequent adaptive adjustments in plateau areas. This pre-training process can be completed on a regular server, with relatively low computational costs, and the pre-trained model can be reused, reducing the initial cost of applications in plateau areas.
[0022] Step S20: Collect the construction risk dataset of new energy power plants in plateau areas as the target domain data.
[0023] In this embodiment, considering the construction characteristics of new energy power plants in plateau regions (such as the Qinghai-Tibet Plateau and the Yunnan-Guizhou Plateau), construction risk data for plateau regions is collected through on-site monitoring systems, construction logs, and historical accident reports, serving as the target domain data for transfer learning. Due to the short construction cycle in plateau regions (typically only 3-6 months), the amount of target domain data is limited, usually between 1000-5000 records. During the data collection process, special attention is paid to risk factors unique to plateaus, such as the thin air caused by high altitude, the decline in equipment performance under low temperatures, and the impact of strong ultraviolet radiation on construction personnel. The collected data needs to be cleaned and standardized to ensure data quality. The difficulty in data collection in plateau regions lies in the tight timeframe and harsh environment; therefore, this solution particularly emphasizes efficient data acquisition and preprocessing to provide high-quality target domain data for subsequent model adaptive adjustments.
[0024] Step S30: Calculate the maximum mean difference between the data distributions of the source domain data and the target domain data.
[0025] In this embodiment, Maximum Mean Discrepancy (MMD) is used as an indicator to measure the difference in distribution between the source and target domain data. MMD is a non-parametric method for measuring distribution distance. It calculates the distance between the mean vectors of two distributions by mapping the data to a regenerative kernel Hilbert space (RKHS). In practice, a Gaussian kernel function can be used to calculate MMD. After obtaining the MMD, it is compared with a preset threshold (e.g., 0.1) to determine whether to use transfer learning or fine-tuning. The computational complexity of MMD is O(mn). In practical applications, sampling or approximation methods can be used to reduce computational costs, ensuring rapid completion even in areas with scarce data, such as high-altitude regions. As a measure of distribution difference, MMD objectively reflects the degree of difference in construction risk data between high-altitude and plain areas, providing a scientific basis for subsequent decision-making.
[0026] Step S40: If the maximum mean difference is greater than a preset threshold, then use the target domain data to perform transfer learning on the pre-trained model to obtain a risk warning model for the plateau region.
[0027] In this embodiment, when the calculated maximum mean difference exceeds a preset threshold (e.g., 0.1), it indicates a significant difference in data distribution between plateau and plain areas, necessitating transfer learning to enhance the model's adaptability to the plateau environment. The principle of transfer learning is to utilize the abundant data in the source domain and, through domain adaptation techniques, enable the model to handle the unique distribution of the target domain, making it particularly suitable for situations where target domain data is scarce. Through this transfer learning, the model can effectively leverage historical experience from plain areas to adapt to the unique environment of plateau regions, significantly improving the accuracy and adaptability of risk warnings.
[0028] In step S50, if the maximum mean difference is not greater than a preset threshold, the pre-trained model is fine-tuned using target domain data to obtain a risk warning model for plateau regions.
[0029] In this embodiment, when the maximum mean difference is no greater than a preset threshold (e.g., 0.1), it indicates that the data distribution difference between the plateau region and the plain region is small, and the pre-trained model can be directly fine-tuned to retain the pre-training advantages of the source domain. The principle of fine-tuning is to adjust only the local parameters of the model, avoiding damage to the overall structure of the pre-trained model, preserving the generalization ability brought by the source domain data, and making targeted adjustments using a small amount of data from the plateau region. This method can quickly obtain a high-performance early warning model when the target domain and the source domain are similarly distributed, and has the advantages of high computational efficiency and short training time.
[0030] In this embodiment, a construction risk dataset of new energy power plants in plain areas (such as the North China Plain and the middle and lower reaches of the Yangtze River Plain) is first obtained from a historical database as the source domain data for transfer learning. This dataset contains various risk factors during construction (such as weather conditions, equipment failure rates, and personnel operational errors) and their corresponding risk levels. Then, the random forest algorithm model is pre-trained using this dataset. By adjusting hyperparameters such as the number of trees and the maximum depth, a basic model with strong generalization ability is constructed, i.e., the pre-trained model. This pre-training process can be completed on a regular server, with relatively low computational costs, and the pre-trained model can be reused, reducing the initial cost of application in plateau areas. Considering the construction characteristics of new energy power plants in plateau areas (such as the Qinghai-Tibet Plateau and the Yunnan-Guizhou Plateau), construction risk data for plateau areas is collected through on-site monitoring systems, construction logs, and historical accident reports, serving as the target domain data for transfer learning. The Maximum Mean Difference (MMD) is used as an indicator to measure the difference in distribution between the source and target domain data. When the calculated MMD exceeds a preset threshold (e.g., 0.1), it indicates a significant difference in data distribution between plateau and plain areas, requiring transfer learning to enhance the model's adaptability to the plateau environment. Through this transfer learning, the model can effectively utilize historical experience from plain areas to adapt to the unique environment of plateau regions, significantly improving the accuracy and adaptability of risk warnings. When the MMD is not greater than the preset threshold (e.g., 0.1), it indicates a smaller difference in data distribution between plateau and plain areas, allowing for direct fine-tuning of the pre-trained model while preserving the advantages of pre-training from the source domain. This method can quickly obtain high-performance warning models when the target and source domains have similar distributions, offering advantages such as high computational efficiency and short training time.
[0031] Furthermore, in one embodiment, the step of using target domain data to perform transfer learning on the pre-trained model to obtain a risk warning model for plateau regions includes: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
[0032] In this embodiment, transfer learning guides the model to reduce the distributional difference between the source and target domains during training by adding a maximum mean difference term to the loss function and incorporating transfer learning weights λ. Specifically, the current loss function of the pre-trained model is updated based on Formula 1 to obtain the first loss function. Then, using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions. During training, the model focuses not only on prediction accuracy but also on the alignment of the source and target domain distributions, thereby better adapting to the construction risk characteristics of plateau regions.
[0033] Furthermore, in one embodiment, the method for early warning of construction risks in plateau areas further includes: Calculate the new maximum mean difference between the data distributions of the source domain data and the target domain data at a preset period; The transfer learning weight λ is increased or decreased accordingly based on whether the new maximum mean difference increases or decreases.
[0034] In this embodiment, the transfer learning weight λ is dynamically adjusted to allow the model to continuously optimize as data from the plateau region accumulates. Specifically, the new maximum mean difference between the source and target domain data is calculated at a preset period (e.g., weekly). Based on whether this new maximum mean difference increases or decreases, the transfer learning weight λ is increased or decreased accordingly. For example, when the maximum mean difference increases, it indicates a greater difference in the distribution between the plateau and plain regions, requiring an increase in the transfer learning weight λ to strengthen domain adaptation; when the maximum mean difference decreases, it indicates a smaller distribution difference, requiring a decrease in the transfer learning weight λ to retain the pre-training advantage of the source domain. This dynamic adjustment mechanism enables the model to adapt to changes in plateau region data and continuously optimize early warning accuracy.
[0035] Furthermore, in one embodiment, the larger the amount of data in the target domain, the smaller the corresponding transfer learning weight λ is set.
[0036] In this embodiment, the larger the amount of data in the target domain, the smaller the corresponding transfer learning weight λ is set. This is because as the amount of data in the target domain increases, the dominant role of the target domain data in the model strengthens, while the relative importance of the source domain data decreases. Therefore, it is necessary to reduce the intensity of transfer learning to avoid over-reliance on the source domain data. This design allows the model to automatically adjust when the amount of data changes, maintaining optimal adaptability.
[0037] Furthermore, in one embodiment, the step of fine-tuning the pre-trained model using target domain data to obtain a risk warning model for plateau regions includes: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
[0038] In this embodiment, fine-tuning updates the current loss function of the pre-trained model using Formula 2 to obtain the second loss function. α and β are hyperparameters, dynamically set based on the magnitude of the maximum mean difference. For example, when the maximum mean difference is small, the values of α and β are reduced accordingly to retain more source domain pre-training knowledge. Using target domain data, the predicted values of only the leaf nodes of the pre-trained model are retrained based on the second loss function to obtain the risk warning model for plateau regions. The fine-tuning process only retrains the predicted values of the leaf nodes of the pre-trained model, preserving the model's tree structure and feature selection capabilities. The leaf node predicted values of the random forest model reflect the model's response to specific feature combinations. By updating only the leaf nodes, modifications to the overall model structure can be minimized, preserving the generalization ability brought by the source domain data, while utilizing plateau region data for targeted optimization. This fine-tuning method is more efficient and computationally cheaper than transfer learning, making it particularly suitable for situations where data is scarce in plateau regions.
[0039] Secondly, this application also provides a construction risk early warning device for plateau areas.
[0040] In one embodiment, reference is made to Figure 2 , Figure 2 This is a functional module diagram of an embodiment of the high-altitude construction risk early warning device of this application, as shown below. Figure 2 As shown, the construction risk early warning device in plateau areas includes: The acquisition module 10 is used to acquire the construction risk dataset of new energy power plants in the plain area as source domain data, and to use the source domain data to pre-train the random forest algorithm model to obtain a pre-trained model. The data acquisition module 20 is used to collect construction risk datasets for new energy power plants in plateau areas as target domain data. Calculation module 30 is used to calculate the maximum mean difference between the data distributions of the source domain data and the target domain data; The transfer module 40 is used to perform transfer learning on the pre-trained model using target domain data if the maximum mean difference is greater than a preset threshold, so as to obtain a risk warning model for the plateau region. The fine-tuning module 50 is used to fine-tune the pre-trained model using target domain data if the maximum mean difference is not greater than a preset threshold, so as to obtain a risk warning model for plateau areas.
[0041] Furthermore, in one embodiment, the migration module 40 is used for: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
[0042] Furthermore, in one embodiment, the high-altitude construction risk early warning device also includes a weight adjustment module, used for: Calculate the new maximum mean difference between the data distributions of the source domain data and the target domain data at a preset period; The transfer learning weight λ is increased or decreased accordingly based on whether the new maximum mean difference increases or decreases.
[0043] Furthermore, in one embodiment, the larger the amount of data in the target domain, the smaller the corresponding transfer learning weight λ is set.
[0044] Furthermore, in one embodiment, the fine-tuning module 50 is used for: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
[0045] The functions of each module in the above-mentioned high-altitude construction risk early warning device correspond to the steps in the above-mentioned high-altitude construction risk early warning method embodiment, and their functions and implementation processes will not be described in detail here.
[0046] Thirdly, this application provides a construction risk early warning device for high-altitude areas.
[0047] Reference Figure 3 , Figure 3 This is a schematic diagram of the hardware structure of a high-altitude construction risk early warning device involved in an embodiment of this application. In this embodiment, the high-altitude construction risk early warning device may include a processor, a memory, a communication interface, and a communication bus.
[0048] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.
[0049] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the high-altitude construction risk early warning equipment, as well as interfaces used for interconnecting the high-altitude construction risk early warning equipment with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.
[0050] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.
[0051] The processor can be a general-purpose processor, which can call the high-altitude construction risk warning program stored in the memory and execute the high-altitude construction risk warning method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the high-altitude construction risk warning program is called can be referred to the various embodiments of the high-altitude construction risk warning method of this application, and will not be repeated here.
[0052] Those skilled in the art will understand that Figure 3 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0053] Fourthly, embodiments of this application also provide a readable storage medium.
[0054] The present application has a readable storage medium storing a high-altitude construction risk warning program, wherein when the high-altitude construction risk warning program is executed by a processor, it implements the steps of the high-altitude construction risk warning method described above.
[0055] The method for implementing the construction risk early warning procedure in plateau areas can be referred to in the various embodiments of the plateau area construction risk early warning method of this application, and will not be repeated here.
[0056] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0057] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.
[0058] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.
[0059] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.
[0060] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.
[0061] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.
[0062] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.
Claims
1. A method for early warning of construction risks in plateau areas, characterized in that, The construction risk early warning method in plateau areas includes: We obtained a construction risk dataset of new energy power plants in the plains area as the source domain data, and used the source domain data to pre-train the random forest algorithm model to obtain a pre-trained model. Data on construction risks of new energy power plants in plateau areas were collected as the target domain data. Calculate the maximum mean difference between the data distributions of the source domain data and the target domain data; If the maximum mean difference is greater than the preset threshold, the target domain data is used to perform transfer learning on the pre-trained model to obtain a risk warning model for the plateau region. If the maximum mean difference is not greater than the preset threshold, the pre-trained model is fine-tuned using target domain data to obtain a risk warning model for plateau regions.
2. The method for early warning of construction risks in plateau areas as described in claim 1, characterized in that, The method of using target domain data to perform transfer learning on the pre-trained model to obtain a risk warning model for plateau areas includes: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
3. The construction risk early warning method for plateau areas as described in claim 2, characterized in that, The method for early warning of construction risks in plateau areas also includes: Calculate the new maximum mean difference between the data distributions of the source domain data and the target domain data at a preset period; The transfer learning weight λ is increased or decreased accordingly based on whether the new maximum mean difference increases or decreases.
4. The construction risk early warning method for plateau areas as described in claim 2, characterized in that, The larger the amount of data in the target domain, the smaller the corresponding transfer learning weight λ should be.
5. The method for early warning of construction risks in plateau areas as described in claim 1, characterized in that, The process of fine-tuning the pre-trained model using target domain data to obtain a risk warning model for plateau regions includes: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
6. A construction risk early warning device for plateau regions, characterized in that, The high-altitude area construction risk early warning device includes: The acquisition module is used to acquire the construction risk dataset of new energy power plants in the plains area as the source domain data, and to use the source domain data to pre-train the random forest algorithm model to obtain the pre-trained model. The data acquisition module is used to collect construction risk datasets for new energy power plants in plateau areas as target domain data. The calculation module is used to calculate the maximum mean difference between the data distributions of the source domain data and the target domain data; The transfer module is used to perform transfer learning on the pre-trained model using target domain data if the maximum mean difference is greater than a preset threshold, so as to obtain a risk warning model for plateau areas. The fine-tuning module is used to fine-tune the pre-trained model using target domain data if the maximum mean difference is not greater than a preset threshold, thereby obtaining a risk warning model for plateau regions.
7. The high-altitude construction risk early warning device as described in claim 6, characterized in that, The migration module is used for: Based on the maximum mean difference, the current loss function of the pre-trained model is updated using Formula 1 to obtain the first loss function, which is: First loss function = current loss function of pre-trained model + λ·maximum mean difference; Where λ is the transfer learning weight; Using target domain data, the pre-trained model is trained based on the first loss function to obtain a risk warning model for plateau regions.
8. The high-altitude construction risk early warning device as described in claim 6, characterized in that, The fine-tuning module is used for: The second loss function is obtained by updating the current loss function of the pre-trained model using Formula 2. Formula 2 is as follows: Second loss function = current loss function of pre-trained model + α·adversarial loss function + β·covariance alignment loss function; Where α and β are hyperparameters, set based on the maximum mean difference; Using target domain data, a risk warning model for plateau regions is obtained by retraining only the predicted values of the leaf nodes of the pre-trained model based on the second loss function.
9. A construction risk early warning device for high-altitude areas, characterized in that, The plateau construction risk early warning device includes a processor, a memory, and a plateau construction risk early warning program stored in the memory and executable by the processor, wherein when the plateau construction risk early warning program is executed by the processor, it implements the steps of the plateau construction risk early warning method as described in any one of claims 1 to 5.
10. A readable storage medium, characterized in that, The readable storage medium stores a high-altitude construction risk warning program, wherein when the high-altitude construction risk warning program is executed by the processor, it implements the steps of the high-altitude construction risk warning method as described in any one of claims 1 to 5.