Work and rest suggestion generation method, system and equipment for electric power operating personnel and medium

Abstract

The invention relates to the technical field of information management, and discloses a work and rest suggestion generation method, system and device for electric power workers and a medium, and the method comprises the steps: obtaining triple data and work and rest data of each work order under a target work order type; triad data and work and rest data of the work order are used as training samples, and a work and rest model suitable for a target work order type is constructed and trained; processing the target triple data of the target work order of the target work order type to which the target work order belongs by adopting a work and rest model to obtain a target work and rest data predicted value; and on the basis of the target work and rest data prediction value, generating a work and rest suggestion of an operator corresponding to the target work order. According to the scheme, a reasonable rest arrangement suggestion can be given after the worker completes the work order, it is guaranteed that the worker recovers to the target health state, and then potential safety hazards of the worker in the subsequent electric power work execution process are reduced.

Description

Technical Field

[0001] This invention relates to the field of information management technology, and in particular to a method, system, device, and medium for generating work and rest suggestions for power workers. Background Technology

[0002] At present, with the rapid development of the power industry, the power working environment has become increasingly complex. This complexity leads to health and safety issues for frontline personnel working in the power system, especially those in the power grid industry. The workload is not easy, so long hours and high intensity work can cause physical and mental exhaustion for frontline workers. The increasingly complex working environment has a deeper impact on the mental health of frontline personnel.

[0003] If frontline workers do not receive adequate rest and recovery after each work session, this can lead to health problems in the long run. This is especially true when workers are assigned to the next work order without proper rest and recovery, which further increases the risk of safety and health hazards during work. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method, system, device and medium for generating work and rest suggestions for power workers, which can solve the problem of health and safety hazards caused by power workers not being able to get reasonable and sufficient rest in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for generating work and rest suggestions for power workers, comprising: For multiple completed work orders of the same target work order type, obtain the triplet data and work schedule data for each work order; The triplet data includes the worker's first health indicator value before the work order is completed, the work process data during the work order, and the second health indicator value after the work order is completed. The work and rest data includes the work and rest content experienced by the workers as they recover from the second health indicator value to the target health indicator value; Using the triplet data of work orders and work schedule data as training samples, a work schedule model suitable for the target work order type is constructed and trained. The input to the work-rest model is triple data, and the output is the predicted value of the sample work-rest data; For the target triplet data of the target work order belonging to the aforementioned target work order type, the work schedule model is used for processing to obtain the predicted value of the target work schedule data. Based on the predicted values ​​of the target work schedule data, work schedule suggestions are generated for the operators corresponding to the target work order.

[0007] As a preferred embodiment of the method for generating work and rest suggestions for power workers according to the present invention, the step of constructing and training a work and rest model suitable for the target work order type using triplet data and work and rest data of work orders as training samples includes: The triplet data in the training samples are input into a pre-built neural network model for processing, and the predicted values ​​of the sample's daily routine data are output. The daily routine data in the training samples are used as the measured values ​​of the sample daily routine data. A loss function is constructed based on the measured values ​​and predicted values ​​of the sample daily routine data. The neural network model is then trained based on the loss function to obtain the daily routine model.

[0008] As a preferred embodiment of the method for generating work and rest suggestions for power workers according to the present invention, the work and rest data includes multiple work and rest items and arrangement information corresponding to the multiple work and rest items, wherein the multiple work and rest items include at least fixed basic work and rest items, or further include multiple additional work and rest items.

[0009] As a preferred embodiment of the method for generating work and rest suggestions for power workers according to the present invention, the step of inputting the triplet data from the training samples into a pre-built neural network model for processing and outputting predicted values ​​of the sample work and rest data includes: The training samples are grouped according to the type of daily routine items in the included daily routine data, resulting in the following multiple data groups: a basic data group containing only basic daily routine items, multiple additional data groups containing both basic and additional daily routine items, where the additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in different additional data groups. The neural network model includes multiple parallel neural network branches that correspond one-to-one with the multiple data groups. For each data set, the triplet data in the data set is input into the corresponding neural network branch for processing, and the predicted value of the sample daily routine data branch is output. The step of using the daily routine data in the training samples as the measured values ​​of the sample daily routine data, constructing a loss function based on the measured values ​​and predicted values ​​of the sample daily routine data, and training the neural network model based on the loss function to obtain the daily routine model includes: For each neural network branch, the daily routine data in the corresponding data set is used as the measured value of the sample daily routine data branch. A branch loss function is constructed based on the measured value and the predicted value of the sample daily routine data branch. The current neural network branch is then trained based on the branch loss function to obtain the daily routine network branch. All daily routine network branches are combined to form the daily routine model.

[0010] This preferred approach improves the accuracy and relevance of work and rest recommendations. By grouping training samples according to work and rest item types and processing them using multiple parallel neural network branches, each branch focusing on a specific type of work and rest data, the model can learn the patterns under different combinations of work and rest items more meticulously. This allows for more accurate predictions of work and rest schedules when processing actual data on power workers' schedules.

[0011] As a preferred embodiment of the method for generating work schedule suggestions for power workers according to the present invention, wherein: the target triplet data of the target work order belonging to the target work order type is processed using the work schedule model to obtain the predicted value of the target work schedule data, including: The target triplet data of the target work order belonging to the target work order type are processed by each work schedule network branch in the work schedule model to obtain multiple target work schedule data branch prediction values. Based on the predicted values ​​of the multiple target daily routine data branches, the predicted value of the target daily routine data is obtained.

[0012] As a preferred embodiment of the method for generating work and rest suggestions for power workers according to the present invention, the step of inputting the triplet data from the training samples into a pre-built neural network model for processing and outputting predicted values ​​of the sample work and rest data includes: The training samples are grouped according to the type of daily routine items in the included daily routine data, resulting in the following multiple data groups: a basic data group containing only basic daily routine items, and multiple additional data groups containing both basic and additional daily routine items. The additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in the data groups between different additional data groups. The neural network model includes a serially connected basic network branch and an additional network branch. The triplet data in the data set is input into the basic network branch for processing, and the basic branch prediction value of the sample daily life data is output. The basic branch prediction value of the sample daily routine data and the given additional daily routine items are input into the additional network branch, and the additional branch prediction value of the sample daily routine data is output.

[0013] As a preferred embodiment of the method for generating work schedule suggestions for power workers according to the present invention, wherein: the target triplet data of the target work order belonging to the target work order type is processed using the work schedule model to obtain the predicted value of the target work schedule data, including: The target triplet data of the target work order belonging to the target work order type is input into the basic branch of the work schedule network, and the given target additional work schedule item is input into the additional branch of the work schedule network for processing to obtain the target work schedule data additional prediction value, and the target work schedule data additional prediction value is used as the target work schedule data prediction value. The step of generating work schedule suggestions for the workers corresponding to the target work order based on the predicted values ​​of the target work schedule data includes: The predicted values ​​of the target work schedule data are optimized and adjusted, and the optimized and adjusted predicted values ​​of the target work schedule data are used as work schedule suggestions for the operators corresponding to the target work order.

[0014] Secondly, the present invention provides a system for generating work and rest suggestions for power workers, comprising: The data acquisition module is used to acquire triplet data and work schedule data for each of the multiple completed work orders belonging to the same target work order type. The triplet data includes the worker's first health indicator value before the work order is completed, the work process data during the work order, and the second health indicator value after the work order is completed. The work and rest data includes the work and rest content experienced by the workers as they recover from the second health indicator value to the target health indicator value; The work schedule model building module is used to construct and train a work schedule model suitable for the target work order type using the triplet data and work schedule data of the work order as training samples. The input to the work-rest model is triple data, and the output is the predicted value of the sample work-rest data; The prediction module is used to process the target triplet data of the target work order belonging to the target work order type using the work schedule model to obtain the predicted value of the target work schedule data. The suggestion module is used to generate work schedule suggestions for the operators corresponding to the target work order based on the predicted values ​​of the target work schedule data.

[0015] Thirdly, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0016] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0017] Compared with existing technologies, the beneficial effect of this invention is that it proposes a method for generating rest and work suggestions for power workers. This method involves acquiring triplet data and rest and work data for each work order under a target work order type; using the triplet data and rest and work data as training samples, a rest and work model suitable for the target work order type is constructed and trained; for the target triplet data of the target work order of the corresponding target work order type, the rest and work model is used to process the target rest and work data prediction value; based on the target rest and work data prediction value, rest and work suggestions for the corresponding workers are generated. This solution can provide reasonable rest and work arrangement suggestions for workers after they complete a work order, ensuring that workers recover to a target health state, thereby reducing safety hazards during subsequent power work. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a method for generating work and rest suggestions for power workers, as provided in one embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of another method for generating work and rest suggestions for power workers, provided as an embodiment of the present invention.

[0021] Figure 3 This is a flowchart illustrating a method for training a work schedule model to generate work schedule suggestions for power workers, as provided in an embodiment of the present invention.

[0022] Figure 4 This is a flowchart illustrating another method for training a work schedule model in a method for generating work schedule suggestions for power workers, as provided in one embodiment of the present invention.

[0023] Figure 5 This is an internal structural diagram of an electronic device for generating work and rest suggestions for power workers, as provided in one embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figures 1-3 This is the first embodiment of the present invention, which provides a method for generating work and rest suggestions for power workers, including: like Figure 1 As shown, the method for generating work and rest suggestions for power workers includes the following steps.

[0026] Step 110: For multiple completed work orders of the same target work order type, obtain the triplet data and work-rest data for each work order. The triplet data includes the worker's first health indicator value before starting the work order, the work process data during the work order, and the second health indicator value after the work order is completed. The work-rest data includes the work-rest content experienced by the worker as they recover from the second health indicator value to the target health indicator value.

[0027] Work order types can be categorized based on multiple dimensions, including the type of work (transmission, substation, distribution, etc.), work content, technical elements, and required skills. To accurately classify work orders, this implementation will simultaneously use all of these dimensions to categorize them. The target work order type can be any existing work order type. For any work order of any given type, the work order can be divided into three stages according to its progress: before work order execution, during work order execution, and after work order completion. Status data of the workers will be collected during these three stages: the first health indicator value before work order execution, the work process data during work order execution, and the second health indicator value after work order completion.

[0028] Health indicator values ​​refer to health information reflecting the physical health of workers, which can be collected through wearable devices, image acquisition devices, and other facilities. This includes health information such as organ function, mental state, and basic vital signs. Those skilled in the art can set health indicator items according to the actual equipment capabilities and needs; these will not be listed here. To differentiate health indicator values ​​collected at different stages, this embodiment records the health indicator values ​​of workers before the work order is completed as the "first health indicator value," and the health indicator values ​​of workers after the work order is completed as the "second health indicator value."

[0029] Work process data refers to the relevant data involved by workers during the work order process, which may include, but is not limited to, the health indicators of workers, work order-related data, and environmental data of the environment in which workers are located.

[0030] The work order-related data includes the work order type, work content, technical elements, and responsibilities / skills.

[0031] Environmental data refers to descriptive data about the environment in which workers are located, such as time period (daytime, nighttime, etc.), geography (surrounding facilities overview, size and / or enclosure of the work environment, altitude, etc.), weather conditions (wind, rain, temperature, humidity, seasonal changes, etc.), noise, light, etc.

[0032] It should be noted that since work order operation is a continuous process, when generating operation process data, the most prominent and obvious representative data collected during the work order operation process, or the data collected at specific sampling points, can be combined to form operation process data through sampling.

[0033] Work-rest data refers to the work-rest activities experienced by workers during the period after a work order is completed, during which their health indicators recover from the second health indicator value to the target health indicator value. This work-rest content includes specific arrangements for various work-rest items such as eating, sleeping, entertainment, and fitness, such as the combination of meals, the time period and duration of sleep, the content and arrangement of entertainment activities, the content and arrangement of fitness activities, and the total duration of work-rest activities.

[0034] In an optional implementation, for multiple completed work orders of the same target work order type, the first health indicator value of the operator before the work order, the work process data during the work order, and the second health indicator value after the work order are collected and summarized to obtain the triplet data for the corresponding work order. The work and rest content experienced by the operator from the second health indicator value after the work order is completed back to the target health indicator value is also collected and summarized to obtain the work and rest data for the corresponding work order. The triplet data and work and rest data for each work order basically represent the complete lifecycle of various state changes experienced by the operator when executing a certain work order.

[0035] Step 120: Using the triplet data and work schedule data of the work order as training samples, construct and train a work schedule model suitable for the target work order type. The input of the work schedule model is the triplet data, and the output is the predicted value of the sample work schedule data.

[0036] Throughout the complete lifecycle of a worker's various state changes during the execution of a work order, the relevant states before, during, and after the work order are related to the required rest and work schedule (the rest and work arrangement before returning to the target health indicator value). For example, in the simplest terms, the higher the intensity of the work order, the more complex the work environment, and the worse the worker's health indicator values ​​are before and after the work order, then the worker will inevitably need more thorough, reasonable, and even prolonged rest and work adjustments after the work order to return to the normal target health indicator value. Theoretically, as long as the rest and work adjustment duration is increased indefinitely, the physical health state can always be restored to the target value. However, in reality, this recovery period cannot be extended indefinitely. Therefore, a reasonable prediction scheme is needed to learn the changing patterns between the triplet data and rest and work schedule data corresponding to the same type of work order, in order to accurately assess which triplet data needs to be paired with which rest and work schedule data to restore to the target health indicator state.

[0037] In one optional implementation, triplet data and work / rest data corresponding to multiple work orders of the same target work order type are used as training samples. The triplet data in the training samples serve as the input to the model to be trained, and the work / rest data serves as the label corresponding to the output predicted value. A work / rest model suitable for the target work order type is constructed and trained using the training samples. The input of this work / rest model is triplet data, and the output is the predicted work / rest data. To distinguish it from other work / rest data, this embodiment refers to the predicted work / rest data obtained based on the training samples during the work / rest model training process as the "sample work / rest data predicted value".

[0038] In some embodiments, such as Figure 2 As shown, in step 120, the process of constructing and training a work schedule model suitable for the target work order type using the triplet data and work schedule data of the work order as training samples may include the following steps.

[0039] Step 210: Input the triplet data from the training samples into the pre-built neural network model for processing, and output the predicted values ​​of the sample daily routine data.

[0040] Regarding the network structure of the neural network model, this embodiment does not limit it and may include, but is not limited to, any one of the following: fully connected neural network (FCN), convolutional neural network (CNN), and residual network (ResNet).

[0041] In an alternative implementation, for training samples composed of triplet data and daily routine data, the triplet data can be input into the pre-built neural network model described above for processing, and the predicted values ​​of the sample daily routine data can be output.

[0042] Step 220: Use the daily routine data in the training samples as the measured values ​​of the sample daily routine data, construct a loss function based on the measured values ​​and predicted values ​​of the sample daily routine data, and train the neural network model based on the loss function to obtain the daily routine model.

[0043] Specifically, during the training process, the work and rest data in the training samples can be used as the measured values ​​of the sample work and rest data, i.e., the label values. A loss function is constructed based on the difference between the measured values ​​of the sample work and rest data and the corresponding predicted values ​​of the above sample work and rest data. The neural network model is then trained based on the loss function, and the network parameters of the neural network model are adjusted. Thus, after the training process is completed, the work and rest model of the target work order type is obtained.

[0044] The form of the loss function is not limited and may include, but is not limited to, squared loss function, absolute loss function, etc.

[0045] In some embodiments, the loss function may include: Where Loss is the loss value of the loss function, x i Let i be the predicted value of the daily routine data for the i-th sample. Let be the measured value of the daily routine data of the i-th sample, and n be the number of samples.

[0046] Step 130: For the target triplet data of the target work order of the corresponding target work order type, use the work schedule model to process it and obtain the predicted value of the target work schedule data.

[0047] In an optional implementation, for a given target triplet data of a target work order belonging to a specific target work order type, the target work order could be a work order that has just finished and the workers have not yet adjusted their work schedules, or a work order that was completed a long time ago. The work schedule model that has been trained and is applicable to the target work order type is used to process the target triplet data to obtain predicted work schedule data. To distinguish it from other predicted work schedule data, this embodiment refers to the predicted work schedule data obtained based on the given target triplet data during the model application process as the "target work schedule data prediction value".

[0048] Step 140: Based on the predicted values ​​of the target work schedule data, generate work schedule suggestions for the operators corresponding to the target work order.

[0049] In one optional implementation, based on the predicted target work-rest data, it is possible to roughly understand the work-rest arrangements required for an operator to recover from the second health indicator value after completing a work order to the target health indicator value. Then, work-rest recommendations are formulated for the operator corresponding to the target work order to help the operator recover their physical health as soon as possible and effectively.

[0050] In some embodiments, when generating work schedule suggestions for workers corresponding to a target work order based on the predicted value of the target work schedule data, in order to quickly provide work schedule suggestions, the predicted value of the target work schedule data can be directly optimized and adjusted, and the optimized and adjusted predicted value of the target work schedule data can be used as the work schedule suggestion for workers corresponding to the target work order.

[0051] One way to optimize and adjust the work schedule is to make appropriate adjustments to the content, order, and duration of the work schedule items in the work schedule data, so as to obtain work schedule suggestions that are more suitable for the current workers.

[0052] The method for generating work and rest suggestions for power workers provided in this implementation obtains triplet data and work and rest data for each of the multiple completed work orders of the target work order type. The triplet data includes the worker's first health indicator value before starting the work order, the work process data during the work order, and the second health indicator value after the work order is completed. The work and rest data includes the work and rest content experienced by the worker from the second health indicator value to the target health indicator value. Using the triplet data and work and rest data of the work orders as training samples, a work and rest model suitable for the target work order type is constructed and trained. The input of the work and rest model is the triplet data, and the output is the predicted value of the sample work and rest data. For the target triplet data of the target work order of the target work order type, the work and rest model is used to process it to obtain the predicted value of the target work and rest data. Based on the predicted value of the target work and rest data, work and rest suggestions for the workers corresponding to the target work order are generated. This solution utilizes triplet data to provide a clearer and more accurate overall description of the changes in the physical health and work process experienced by workers before, during, and after a target work order type. It then learns the relationship between these changes and the rest and activity data required for workers to recover to their target health state after the work order ends—this is the rest and activity model. The model predicts the target triplet data for a given target work order type and generates rest and activity recommendations based on the predicted data. This ensures workers effectively recover to their target health state, thereby reducing safety hazards during subsequent electrical work.

[0053] Based on the above embodiments, this embodiment further optimizes the method for generating work and rest suggestions for power workers. The aforementioned work and rest data may include multiple work and rest items and the arrangement information corresponding to the multiple work and rest items. The multiple work and rest items include at least fixed basic work and rest items, or may further include multiple additional work and rest items.

[0054] The daily routine items can be, but are not limited to, different types of daily routine items such as eating, sleeping, entertainment, and fitness. The arrangement information corresponding to multiple daily routine items can include the specific arrangements for each daily routine item individually, such as the food combination for meals, the time period and duration of sleep, the content and arrangement of entertainment activities, the content and arrangement of fitness activities, and can also include the overall arrangement between multiple daily routine items, such as the overall duration of these daily routine items, the combination relationship between each daily routine item, and precautions.

[0055] Each daily routine data set may include multiple routine items, among which at least one type of fixed routine item is included, such as eating and sleeping. These basic routine items are essential to the daily routine, and in this embodiment, they are referred to as "basic routine items." Each routine data set will contain basic routine items, and the number and type of basic routine items in each routine data set are fixed and identical. Furthermore, each routine data set may also include a type of optional routine items, such as a certain sport (playing ball, swimming), entertainment, recuperation, and rehabilitation training programs. These are non-essential routine items that are flexibly added according to actual needs and supply conditions, and in this embodiment, they are referred to as "additional routine items." A routine data set may contain only basic routine items, or it may contain both basic and additional routine items. The additional routine items included in different routine data sets are not entirely the same; they may differ in number or type.

[0056] It should be noted that the types of daily routine items corresponding to the basic routine items, as well as the types of daily routine items corresponding to the supplementary routine items, can be flexibly adjusted according to changes in quality of life and living conditions. The basic routine items and supplementary routine items can also be different at different times. For example, the same type of routine item may be used as a basic routine item at one time and as a supplementary routine item at another time.

[0057] Accordingly, in the process of executing step 120, this embodiment provides a specific implementation method for training a work-rest model based on a pre-built neural network model, namely, a training method for the work-rest model.

[0058] like Figure 3 As shown in the figure, the training method of the daily routine model provided in this embodiment includes the following steps.

[0059] Step 310: Group the training samples according to the type of daily routine items in the included daily routine data to obtain the following multiple data groups: a basic data group containing only basic daily routine items, multiple additional data groups containing both basic and additional daily routine items, where the additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in different additional data groups, and the neural network model includes multiple parallel neural network branches that correspond one-to-one with the multiple data groups.

[0060] In an optional implementation, in order to learn the classification patterns between triplet data and daily routine data under various combinations of daily routine items, the training samples can be first grouped according to the type set of daily routine items contained in the daily routine data as follows.

[0061] The basic data set is the set of training samples corresponding to the daily routine data that only contain the basic daily routine items. Since all daily routine data contain the same basic daily routine items, there is only one basic data set.

[0062] The supplementary data set is a collection of training samples corresponding to daily routine data that includes both basic and supplementary routine items. Since the types of supplementary routine items may differ across different daily routine data sets, training samples corresponding to daily routine data with the same set of supplementary routine item types are grouped together. This results in multiple supplementary data sets. Within the same supplementary data set, the supplementary routine items are identical across all daily routine data sets, while the supplementary routine items are not entirely identical across different supplementary data sets.

[0063] To better learn the variation patterns between triplet data and daily routine data within different data sets, the neural network model can be pre-built with multiple parallel neural network branches. These parallel branches correspond one-to-one with the multiple data sets formed after all training samples are divided. The network structures of these parallel branches can be identical or not.

[0064] Step 320: For each data group, input the triplet data in the data group into the corresponding neural network branch for processing, and output the sample daily routine data branch prediction value.

[0065] Specifically, for each data set, the triplet data and daily routine data in that data set are used as training samples. The triplet data in the training samples serve as the input to the corresponding neural network branch, and the daily routine data serves as the label corresponding to the output predicted value. A neural network branch suitable for that data set is constructed and trained using the training samples. The input of this neural network branch is the triplet data, and the output is the predicted daily routine data. To distinguish it from other daily routine data, this embodiment refers to the predicted daily routine data obtained based on the training samples during the training process of each neural network branch as the "sample daily routine data branch predicted value." It is evident that the daily routine prediction effects expressed by the predicted values ​​obtained after processing by different neural network branches are different, the difference being reflected in the different combinations of the types of daily routine items in the daily routine data.

[0066] Steps 310-320 above can be used as a detailed process of inputting the triplet data in the training samples into the pre-built neural network model for processing and outputting the predicted values ​​of the sample daily routine data in step 210 of the aforementioned embodiment.

[0067] Step 330: For each neural network branch, take the daily routine data in the corresponding data group as the measured value of the sample daily routine data branch, and construct a branch loss function based on the measured value and the predicted value of the sample daily routine data branch. Then, train the current neural network branch based on the branch loss function to obtain the daily routine network branch. All daily routine network branches are combined to form the daily routine model.

[0068] In one optional implementation, during the training of each neural network branch, the daily routine data in the data set (training samples) corresponding to each neural network branch can be used as the measured value of the sample daily routine data branch, i.e., the label value. Based on the difference between the measured value of the sample daily routine data branch and the corresponding predicted value of the aforementioned sample daily routine data branch, a branch loss function is constructed for that neural network branch. The neural network branch is then trained based on the branch loss function, and the network parameters of the neural network branch are adjusted. Thus, after the training process is completed, a daily routine network branch corresponding to the set of daily routine items in each data set is obtained. All daily routine network branches are combined to form the aforementioned daily routine model.

[0069] The form of the loss function for each branch is not limited and may include, but is not limited to, the squared loss function and the absolute loss function.

[0070] In some embodiments, the branch loss function may include: Where Loss1 is the loss value of the branch loss function, x 1i Let i be the predicted value of the activity data for the i-th sample. Let m be the measured value of the daily routine data branch for the i-th sample, and m be the number of samples.

[0071] Step 330 above can be used as step 220 in the aforementioned embodiment, whereby the daily routine data in the training samples are used as the measured values ​​of the sample daily routine data, a loss function is constructed based on the measured values ​​and predicted values ​​of the sample daily routine data, and the neural network model is trained based on the loss function to obtain a detailed process of refining the daily routine model.

[0072] Based on this, step 130 above, which processes the target triplet data of the target work order of the corresponding target work order type using a work schedule model to obtain the predicted value of the target work schedule data, may include: Step a: Process the target triplet data of the target work order of the target work order type using the various work schedule network branches in the work schedule model to obtain multiple target work schedule data branch prediction values.

[0073] Specifically, for a given set of target triplet data for a target work order of a given target work order type, this embodiment can predict the work schedule data under different work schedule item type sets. This involves inputting the target triplet data into each of the trained work schedule network branches for processing, thereby obtaining multiple predicted values ​​for work schedule data under different work schedule item type sets. To distinguish it from other work schedule data, this embodiment refers to the predicted work schedule data obtained by each neural network branch based on the given target triplet data of the target work order during application as a "target work schedule data branch prediction value." Thus, each work schedule network branch can output a target work schedule data branch prediction value for the target triplet data, and the difference between these target work schedule data branch prediction values ​​lies in the different combinations of work schedule item types in the work schedule data.

[0074] Step b: Based on the predicted values ​​of multiple target daily routine data branches, obtain the predicted value of the target daily routine data.

[0075] Specifically, after obtaining the predicted values ​​of multiple target daily routine data branches output by the above multiple daily routine network branches, these predicted values ​​can be comprehensively considered to finally obtain a target daily routine data predicted value. For example, any one of the target daily routine data branch predicted values ​​can be directly used as the target daily routine data predicted value.

[0076] In this embodiment, the training samples are grouped according to the types of daily routine items in the included routine data, resulting in multiple data groups: a basic data group containing only basic routine items, and multiple supplementary data groups containing both basic and supplementary routine items. Then, each data group is used to train the corresponding neural network branch, learning the relationship between the triplet data and the routine data in each data group, resulting in multiple routine network branches, and ultimately a routine model composed of these multiple network branches. This routine model can first predict the predicted values ​​of multiple target routine data branches under different routine item type sets for a given target triplet data, and then comprehensively consider these multiple target routine data branch prediction values ​​to obtain a single target routine data prediction value. This provides a convenient and fast scheme for constructing a routine model and a prediction scheme for the target routine data prediction value.

[0077] Example 2, refer to Figure 4 Based on the above embodiments, this embodiment further optimizes the method for generating work and rest suggestions for power workers. The aforementioned work and rest data may include multiple work and rest items and corresponding arrangement information for each item. These multiple work and rest items may include at least fixed basic work and rest items, or may further include multiple supplementary work and rest items. For explanations of the work and rest data, basic work and rest items, please refer to the relevant content in Embodiment Two, which will not be repeated here.

[0078] Accordingly, in the process of executing step 120, this embodiment provides another specific implementation method for training a work-rest model based on a pre-built neural network model, namely, a training method for the work-rest model.

[0079] like Figure 4 As shown in the figure, the training method of the daily routine model provided in this embodiment includes the following steps.

[0080] Step 410: Group the training samples according to the type of daily routine items in the included daily routine data to obtain the following multiple data groups: a basic data group containing only basic daily routine items, multiple additional data groups containing both basic and additional daily routine items, where the additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in the data groups between different additional data groups. The neural network model includes a basic network branch and additional network branches connected in series.

[0081] In this step, the grouping process and results of the training samples are the same as those in step 310 above, and will not be repeated here.

[0082] Unlike step 310, the structural design of the neural network model in this step includes a serially connected basic network branch and an additional network branch. As the names suggest, the basic network branch is used to learn the relationship between the triplet data and the daily routine data in the basic data set; the additional network branch learns the relationship between the predicted daily routine data (containing only basic routine items) based on the triplet data in the additional data set and the daily routine data (the daily routine data in the additional data set) that further includes additional routine items. Simply put, the role of the additional network branch is to learn the transformation rules from daily routine data containing only basic routine items to daily routine data that is further expanded to include additional routine items of different types.

[0083] The reason for using serially connected basic and supplementary network branches as the neural network model to be trained is as follows: Under a given target work order type, there is a certain pattern of change between the triplet data of any work order and its corresponding work schedule data containing only basic work schedule items. By learning this pattern, i.e., training the basic network branch, we can predict how to obtain work schedule data containing only basic work schedule items for triplet data under the target work order type. Then, based on the work schedule data containing only basic work schedule items, we can continuously adjust and add different types of supplementary work schedule items to see how the corresponding work schedule data changes. In other words, training the supplementary network branch can predict how to obtain work schedule data containing a certain supplementary work schedule item from the predicted work schedule data containing only basic work schedule items for triplet data under the target work order type.

[0084] Step 420: Input the triplet data in the data set into the basic network branch for processing, and output the basic branch prediction value of the sample daily routine data.

[0085] In an optional implementation, the role of the base network branch is to receive and process the triplet data from the aforementioned data groups, and output the predicted daily routine data. To distinguish it from other daily routine data, this embodiment refers to the predicted daily routine data obtained by the base network branch based on the training samples throughout the entire neural network model training process as the "sample daily routine data base branch prediction value." It can be understood that this sample daily routine data base branch prediction value is the daily routine data predicted for the triplet data, containing only the basic daily routine items. When a training sample containing a certain triplet data belongs to the base data group, the obtained sample daily routine data base branch prediction value is exactly equivalent to the sample daily routine data branch prediction value output by the neural network branch corresponding to the base data group.

[0086] Step 430: Input the basic branch prediction value of the sample daily routine data and the given additional daily routine items into the additional network branch, and output the additional branch prediction value of the sample daily routine data.

[0087] In one optional implementation, the role of the basic network branch is to predict sleep data containing only basic sleep items for triplet data. To predict sleep data further including additional sleep items of different types, the basic network branch's predicted value of the sample sleep data and the given additional sleep items can be input into the additional network branch to predict sleep data containing additional sleep items. The given additional sleep items are limited to those that must appear in the predicted sleep data. To distinguish it from other sleep data, this embodiment refers to the sleep data predicted by the additional network branch based on training samples during the entire neural network model training process as the "sample sleep data additional branch prediction value." It can be understood that this sample sleep data additional branch prediction value is sleep data further including additional sleep items predicted for triplet data. When a training sample containing a triplet data belongs to a certain additional data group, the obtained sample sleep data additional branch prediction value is exactly equivalent to the sample sleep data branch prediction value output by the neural network branch corresponding to that additional data group.

[0088] Steps 410-430 above can be used as a detailed process of inputting the triplet data in the training samples into the pre-built neural network model for processing and outputting the predicted values ​​of the sample daily routine data in step 210 of the aforementioned embodiment.

[0089] Step 440: During the training of the basic network branch, the daily routine data in the basic data group is used as the basic measured value of the sample daily routine data. A basic loss function is constructed based on the basic measured value and the basic predicted value of the sample daily routine data. The basic network branch is then trained based on the basic loss function to obtain the basic branch of the daily routine network.

[0090] In this embodiment, when training a neural network model consisting of basic network branches and additional network branches, the basic network branches are trained first, and then the additional network branches are trained after the trained basic network branches are fixed.

[0091] Specifically, when training the basic network branch, the training samples used are those from the aforementioned basic data set. The triplet data from the training samples are input into the basic network branch to obtain the basic predicted values ​​of the sample's daily routine data. Simultaneously, the daily routine data from the training samples is used as the basic measured values ​​of the sample's daily routine data, i.e., the label values. Based on the difference between the basic measured values ​​of the sample's daily routine data and the corresponding basic predicted values, a basic loss function is constructed for this basic network branch. The basic network branch is then trained based on this basic loss function, adjusting its network parameters. Thus, after the training process, a basic branch of the daily routine network corresponding to the set of types of daily routine items in the basic data set is obtained.

[0092] The form of the basic loss function is not limited and may include, but is not limited to, the squared loss function and the absolute loss function.

[0093] In some embodiments, the basic loss function may include: Where Loss2 is the loss value of the basic loss function, x 2i This is the basic predicted value of the daily routine data for the i-th sample. Let k be the basic measured value of the daily routine data for the i-th sample, and k be the number of samples.

[0094] Step 450: During the training of the additional network branch, the daily routine data in the additional data set is used as the additional measured value of the sample daily routine data. An additional loss function is constructed based on the additional measured value and the additional predicted value of the sample daily routine data. The additional network branch is trained based on the additional loss function to obtain the additional branch of the daily routine network. Among them, the additional predicted value of the sample daily routine data is obtained by processing the sample daily routine data basic branch predicted value output by the basic branch of the daily routine network and the given additional daily routine items through the additional network branch. The basic branch of the daily routine network and the additional branch of the daily routine network are connected in series to form the daily routine model.

[0095] Specifically, before training the additional network branches, the network structure of the already trained basic branch of the sleep-wake network is fixed. Then, training samples belonging to the aforementioned additional data groups are used to train the additional network branches. During training, the triplet data from the training samples are input into the basic network branch to obtain the basic predicted value of the sample sleep-wake data. Next, the basic predicted value of the sample sleep-wake data and the given additional sleep-wake items are input into the additional network branch to obtain the additional branch predicted value of the sample sleep-wake data. The given additional sleep-wake items are those included in the sleep-wake data of the training samples. The sleep-wake data in the training samples is used as the additional measured value of the sample sleep-wake data, i.e., the label value. Based on the difference between the additional measured value of the sample sleep-wake data and the corresponding additional predicted value of the sample sleep-wake data, an additional loss function is constructed for this additional network branch. The additional network branch is trained based on the additional loss function, and the network parameters of the additional network branch are adjusted. Thus, after the training process is completed, the sleep-wake network additional branches corresponding to the type set of sleep-wake items in all additional data groups are obtained.

[0096] The form of the additional loss function is not limited and may include, but is not limited to, squared loss function, absolute loss function, etc.

[0097] In some embodiments, the additional loss function may include: Where Loss3 is the loss value of the additional loss function, x 3i Add a predicted value to the daily routine data of the i-th sample. Add the measured value to the daily routine data of the i-th sample, where t is the number of samples.

[0098] Steps 440-450 above can be used as step 220 in the aforementioned embodiment, where the daily routine data in the training samples are used as the measured values ​​of the sample daily routine data, a loss function is constructed based on the measured values ​​and predicted values ​​of the sample daily routine data, and the neural network model is trained based on the loss function to obtain a detailed process of refining the daily routine model.

[0099] Based on this, step 130 above, which processes the target triplet data of the target work order of the corresponding target work order type using a work schedule model to obtain the predicted value of the target work schedule data, may include: Step c: Input the target triplet data of the target work order of the target work order type into the basic branch of the work schedule network, input the given target additional work schedule item into the additional branch of the work schedule network for processing, obtain the target work schedule data additional prediction value, and use the target work schedule data additional prediction value as the target work schedule data prediction value.

[0100] Specifically, for a given target triplet data of a target work order belonging to a specific target work order type, this embodiment can predict the work schedule data under different work schedule item type sets. That is, the target triplet data is first input into the trained basic branch of the work schedule network for processing to obtain predicted values ​​of work schedule data under type sets containing only basic work schedule items. To distinguish it from other work schedule data, this embodiment refers to the predicted work schedule data obtained by the basic branch of the work schedule network based on the target triplet data of a given target work order during application as the "target work schedule data basic prediction value". Next, the additional branch of the work schedule network processes the target work schedule data basic prediction value and the given target additional work schedule items input externally to obtain predicted values ​​of work schedule data that further include these target additional work schedule items on top of the basic work schedule items. To distinguish it from other work schedule data, this embodiment refers to the predicted work schedule data obtained by the additional branch of the work schedule network based on the target triplet data of a given target work order and the given target additional work schedule items during application as the "target work schedule data additional prediction value". The additional predicted value of the target daily routine data is used as the target daily routine data predicted value output by the entire daily routine model. It can be understood that if the daily routine model detects that the given target additional daily routine item is empty, the daily routine model can directly use the basic predicted value of the target daily routine data as the target daily routine data predicted value.

[0101] In this embodiment, the training samples are grouped according to the types of daily routine items in the included daily routine data, resulting in the following data groups: a basic data group containing only basic daily routine items, and multiple supplementary data groups containing both basic and supplementary daily routine items. The basic data group is used to train the basic network branch, learning the relationship between the triplet data and the daily routine data, thus obtaining the basic branch of the daily routine network. The supplementary data group is used to train the supplementary network branch, learning the transformation relationship from basic data items to daily routine data combining different types of supplementary daily routine items, thus obtaining the supplementary branch of the daily routine network. This results in a daily routine model composed of the basic branch and the supplementary branch. This model can first predict the basic predicted value of the target daily routine data under the type set of basic daily routine items for a given target triplet data, and then combine it with the given target supplementary daily routine items to obtain the supplementary predicted value of the target daily routine data, i.e., the target daily routine data predicted value. This provides a convenient and fast scheme for constructing a daily routine model and a scheme for predicting the target daily routine data predicted value.

[0102] Example 3, referring to Figure 5 This embodiment also provides a system for generating work and rest suggestions for power workers, including: The data acquisition module is used to acquire triplet data and work schedule data for each of the multiple completed work orders belonging to the same target work order type. The triplet data includes the worker's first health indicator value before the work order is completed, the work process data during the work order, and the second health indicator value after the work order is completed. The work-rest data includes the work-rest content experienced by workers as they recover from the second health indicator value to the target health indicator value; The work schedule model building module is used to build and train a work schedule model suitable for the target work order type using the triplet data and work schedule data of the work order as training samples. The input to the daily routine model is triplet data, and the output is the predicted value of the sample daily routine data; The prediction module is used to process the target triplet data of the target work order of the corresponding target work order type using a work schedule model to obtain the predicted value of the target work schedule data. The suggestion module is used to generate work schedule suggestions for the operators corresponding to the target work order based on the predicted values ​​of the target work schedule data.

[0103] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0104] This embodiment also provides an electronic device, which can be a terminal, and its internal structure diagram can be as follows: Figure 5 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for generating work schedule suggestions for power workers. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the device's casing, or an external keyboard, touchpad, or mouse.

[0105] This embodiment also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps: For multiple completed work orders of the same target work order type, obtain the triplet data and work schedule data for each work order; The triplet data includes the worker's first health indicator value before starting the work order, the work process data during the work order, and the second health indicator value after the work order is completed. The work-rest data includes the work-rest content experienced by workers as they recover from the second health indicator value to the target health indicator value; Using the triplet data of work orders and work schedule data as training samples, a work schedule model suitable for the target work order type is constructed and trained. The input to the daily routine model is triplet data, and the output is the predicted value of the sample daily routine data; For the target triplet data of the target work order of the corresponding target work order type, the work schedule model is used to process the data to obtain the predicted value of the target work schedule data; Based on the predicted values ​​of the target work schedule data, work schedule suggestions are generated for the operators corresponding to the target work order.

[0106] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0107] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0108] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for generating work and rest suggestions for power workers, characterized in that, include: For multiple completed work orders of the same target work order type, obtain the triplet data and work schedule data for each work order; The triplet data includes the worker's first health indicator value before the work order is completed, the work process data during the work order, and the second health indicator value after the work order is completed. The work and rest data includes the work and rest content experienced by the workers as they recover from the second health indicator value to the target health indicator value; Using the triplet data of work orders and work schedule data as training samples, a work schedule model suitable for the target work order type is constructed and trained. The input to the work-rest model is triple data, and the output is the predicted value of the sample work-rest data; For the target triplet data of the target work order belonging to the aforementioned target work order type, the work schedule model is used for processing to obtain the predicted value of the target work schedule data. Based on the predicted values ​​of the target work schedule data, work schedule suggestions are generated for the operators corresponding to the target work order.

2. The method for generating work and rest suggestions for power workers as described in claim 1, characterized in that, The step of constructing and training a work schedule model suitable for the target work order type using triplet data of work orders and work schedule data as training samples includes: The triplet data in the training samples are input into a pre-built neural network model for processing, and the predicted values ​​of the sample's daily routine data are output. The daily routine data in the training samples are used as the measured values ​​of the sample daily routine data. A loss function is constructed based on the measured values ​​and predicted values ​​of the sample daily routine data. The neural network model is then trained based on the loss function to obtain the daily routine model.

3. The method for generating work and rest suggestions for power workers as described in claim 2, characterized in that, The schedule data includes multiple schedule items and corresponding arrangement information for the multiple schedule items. The multiple schedule items include at least fixed basic schedule items, or may further include multiple additional schedule items.

4. The method for generating work and rest suggestions for power workers as described in claim 3, characterized in that, The process of inputting triplet data from the training samples into a pre-built neural network model for processing and outputting predicted values ​​for the sample's daily routine data includes: The training samples are grouped according to the type of daily routine items in the included daily routine data, resulting in the following multiple data groups: a basic data group containing only basic daily routine items, multiple additional data groups containing both basic and additional daily routine items, where the additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in different additional data groups. The neural network model includes multiple parallel neural network branches that correspond one-to-one with the multiple data groups. For each data set, the triplet data in the data set is input into the corresponding neural network branch for processing, and the predicted value of the sample daily routine data branch is output. The step of using the daily routine data in the training samples as the measured values ​​of the sample daily routine data, constructing a loss function based on the measured values ​​and predicted values ​​of the sample daily routine data, and training the neural network model based on the loss function to obtain the daily routine model includes: For each neural network branch, the daily routine data in the corresponding data set is used as the measured value of the sample daily routine data branch. A branch loss function is constructed based on the measured value and the predicted value of the sample daily routine data branch. The current neural network branch is then trained based on the branch loss function to obtain the daily routine network branch. All daily routine network branches are combined to form the daily routine model.

5. The method for generating work and rest suggestions for power workers as described in claim 4, characterized in that, The target triplet data for the target work order belonging to the aforementioned target work order type is processed using the work schedule model to obtain the predicted value of the target work schedule data, including: The target triplet data of the target work order belonging to the target work order type are processed by each work schedule network branch in the work schedule model to obtain multiple target work schedule data branch prediction values. Based on the predicted values ​​of the multiple target daily routine data branches, the predicted value of the target daily routine data is obtained.

6. The method for generating work and rest suggestions for power workers as described in claim 5, characterized in that, The process of inputting triplet data from the training samples into a pre-built neural network model for processing and outputting predicted values ​​for the sample's daily routine data includes: The training samples are grouped according to the type of daily routine items in the included daily routine data, resulting in the following multiple data groups: a basic data group containing only basic daily routine items, and multiple additional data groups containing both basic and additional daily routine items. The additional daily routine items are the same in each data group within the same additional data group, and the additional daily routine items are not completely the same in the data groups between different additional data groups. The neural network model includes a serially connected basic network branch and an additional network branch. The triplet data in the data set is input into the basic network branch for processing, and the basic branch prediction value of the sample daily life data is output. The basic branch prediction value of the sample daily routine data and the given additional daily routine items are input into the additional network branch, and the additional branch prediction value of the sample daily routine data is output.

7. The method for generating work and rest suggestions for power workers as described in claim 6, characterized in that, The target triplet data for the target work order belonging to the aforementioned target work order type is processed using the work schedule model to obtain the predicted value of the target work schedule data, including: The target triplet data of the target work order belonging to the target work order type is input into the basic branch of the work schedule network, and the given target additional work schedule item is input into the additional branch of the work schedule network for processing to obtain the target work schedule data additional prediction value, and the target work schedule data additional prediction value is used as the target work schedule data prediction value. The step of generating work schedule suggestions for the workers corresponding to the target work order based on the predicted values ​​of the target work schedule data includes: The predicted values ​​of the target work schedule data are optimized and adjusted, and the optimized and adjusted predicted values ​​of the target work schedule data are used as work schedule suggestions for the operators corresponding to the target work order.

8. A system for generating work and rest suggestions for power workers, using the method described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire triplet data and work schedule data for each of the multiple completed work orders belonging to the same target work order type. The triplet data includes the worker's first health indicator value before the work order is completed, the work process data during the work order, and the second health indicator value after the work order is completed. The work and rest data includes the work and rest content experienced by the workers as they recover from the second health indicator value to the target health indicator value; The work schedule model building module is used to construct and train a work schedule model suitable for the target work order type using the triplet data and work schedule data of the work order as training samples. The input to the work-rest model is triple data, and the output is the predicted value of the sample work-rest data; The prediction module is used to process the target triplet data of the target work order belonging to the target work order type using the work schedule model to obtain the predicted value of the target work schedule data. The suggestion module is used to generate work schedule suggestions for the operators corresponding to the target work order based on the predicted values ​​of the target work schedule data.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for generating work and rest suggestions for power workers according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for generating work and rest suggestions for power workers according to any one of claims 1 to 7.

Images