Multi-data joint prediction method, device, electronic device and storage medium
By adopting a multi-data joint prediction method in the system, using the historical working condition data of timing characteristics for joint training and prediction of working conditions, energy consumption and load models, the redundancy and information utilization of multiple data to be predicted in the prior art are solved. The problem of insufficient achieves efficient and accurate prediction and fault perception.
Patent Information
- Application Number
- CN202210053986.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-18
- Publication Date
- 2024-12-27
- Estimated Expiration
- 2042-01-18
AI Technical Summary
The prior art cannot effectively utilize timing information in the prediction process of multiple data to be predicted in the system, resulting in redundant models and long training and prediction time. The existing working condition prediction method can only use short-term or instantaneous sensor status information, and cannot Use information effectively.
Using multi-data joint prediction method, the historical working condition data with timing characteristics is input into the pre-trained working condition prediction model, energy consumption prediction model and load warning model for joint training and prediction process, and the model is connected in series. In this article, the warning result and prediction value are output based on the relationship between the load prediction value and the preset threshold value, and the model is updated in real time by monitoring the incremental data.
By reducing the redundancy of the model structure and effectively utilizing timing characteristics, the prediction efficiency and accuracy of multiple prediction data are improved, the working conditions and energy consumption can be predicted more accurately, and equipment failures can be sensed in a timely manner, which improves the accuracy and speed of the prediction value.
Smart Images
Figure CN114528754B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of machine learning technology, and in particular to a multi-data joint prediction method, device, electronic device and storage medium. Background Art
[0002] Neural network belongs to the field of machine learning and cognitive science. It is a mathematical model or computational model with bionic structure and function, which is used to estimate or approximate functions. Neural network is calculated by connecting a large number of artificial neurons, has learning function, and can adaptively change its internal structure based on external information. The learning of neural network is the process of correcting the weights of each layer through the input of training samples. After effective learning, the neural network has excellent performance in data prediction tasks.
[0003] At present, for multiple data to be predicted in a system, a model is usually built for each data to be predicted for training and prediction, and discrete data points are used as training data, which cannot effectively utilize the time series information in the data. In addition, multiple data to be predicted correspond to multiple independent models, which are structurally redundant and take a long time to train and predict.
[0004] In addition, existing operating condition prediction / energy consumption prediction methods are mostly based on the drawing of operating condition curves for prediction, and generally can only use short-term or instantaneous sensor status information, and cannot effectively use the information. Summary of the invention
[0005] In order to solve the above technical problems or at least partially solve the above technical problems, the embodiments of the present application provide a multi-data joint prediction method, device, electronic device and storage medium.
[0006] In a first aspect, an embodiment of the present application provides a multi-data joint prediction method, comprising the following steps:
[0007] Inputting historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model;
[0008] Input the historical operating condition data and operating condition prediction data with time series characteristics into the pre-trained energy consumption prediction model to obtain the energy consumption prediction value;
[0009] Inputting historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process;
[0010] According to the size relationship between the load prediction value and the preset load threshold, the load warning result and the operating condition prediction value and the energy consumption prediction value are determined and output.
[0011] In a possible implementation, the method further includes:
[0012] Real-time monitoring of the data volume of current working condition data, energy consumption value and load warning results and whether the increment between the data volume used in the last model training reaches the preset increment threshold:
[0013] When the increment between the amount of data of the current operating condition data, energy consumption value and load warning results and the amount of data used in the last model training reaches the preset increment threshold, the incremental operating condition data, energy consumption value and load warning results are used to jointly update the operating condition prediction model, energy consumption prediction model and load warning model connected in series.
[0014] In a possible implementation, the joint training during the model training process includes:
[0015] Using known historical operating condition data with time series characteristics as input and target operating condition data as output to train the operating condition prediction model, a trained operating condition prediction model is obtained;
[0016] Input known historical operating condition data with time series characteristics into the trained operating condition prediction model to obtain the output value of the operating condition prediction model;
[0017] Taking known historical operating condition data with time series characteristics and the output value of the trained operating condition prediction model as input values, taking known energy consumption values as output values, training the energy consumption prediction model, and obtaining a trained energy consumption prediction model;
[0018] The known historical operating condition data with time series characteristics, the output value of the trained operating condition prediction model and the output value of the trained energy consumption prediction model are used as input values, and the known load prediction value is used as the output value to train the load warning model to obtain a trained load warning model.
[0019] In a possible implementation, the operating condition prediction model is selected from a recurrent neural network, a long short-term memory network, and a threshold recurrent unit network based on historical operating condition data with time series characteristics; the energy consumption prediction model is a multi-layer perceptron network or a convolutional neural network; the load warning model is a convolutional neural network, wherein the initial weight values corresponding to the historical operating condition data with time series characteristics in the operating condition prediction model, the energy consumption prediction model, and the load warning model present a decay characteristic from back to front according to the time of the historical data.
[0020] In a possible implementation manner, the operating condition prediction model is selected from a recurrent neural network, a long short-term memory network, and a threshold recurrent unit network according to historical operating condition data with time series characteristics, and includes:
[0021] During the joint training of the operating condition prediction model, energy consumption prediction model and load warning model connected in series in sequence, when the time interval of the historical operating condition data with time series characteristics is less than a preset time threshold, selecting a recurrent neural network as the operating condition prediction model;
[0022] When the time interval of historical operating condition data with time series characteristics is greater than or equal to the preset time threshold, the long short-term memory network is selected as the operating condition prediction model;
[0023] When the amount of historical operating condition data with time series characteristics is less than a preset data amount threshold and the time interval is greater than or equal to a preset time length threshold, the threshold cycle unit network is selected as the operating condition prediction model.
[0024] In a possible implementation, the method further includes:
[0025] When the amount of historical operating condition data with time series characteristics is less than a preset data amount threshold and the time interval is less than a preset time length threshold, the bootstrap resampling method is used to resample the historical operating condition data with time series characteristics, so as to train the selected recurrent neural network based on the resampled historical operating condition data with time series characteristics.
[0026] In a possible implementation, the method further includes:
[0027] The historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics are input into a pre-trained fault prediction model, and a fault warning result is output, wherein the fault prediction model is connected in series with the energy consumption prediction model so that the input value of the fault prediction model includes the output value of the energy consumption prediction model.
[0028] In a second aspect, an embodiment of the present application provides a multi-data joint prediction device, comprising:
[0029] An operating condition prediction module, which is used to input historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model;
[0030] An energy consumption prediction module is used to input historical operating condition data and operating condition prediction data with time series characteristics into a pre-trained energy consumption prediction model to obtain energy consumption prediction values;
[0031] A load prediction module, which is used to input historical operating condition data with time series characteristics, operating condition prediction data and energy consumption prediction values into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process;
[0032] The output module is used to determine and output the load warning result and the working condition prediction value and energy consumption prediction value according to the size relationship between the load prediction value and the preset load threshold.
[0033] In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;
[0034] Memory, used to store computer programs;
[0035] The processor is used to implement the steps of the above-mentioned multi-data joint prediction method when executing the program stored in the memory.
[0036] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the above-mentioned multi-data joint prediction method are implemented.
[0037] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:
[0038] The method provided by the embodiment of the present application inputs historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model; inputs historical operating condition data and operating condition prediction data with time series characteristics into a pre-trained energy consumption prediction model to obtain energy consumption prediction values; inputs historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics into a pre-trained load warning model to obtain load prediction values, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process; according to the size relationship between the load prediction value and the preset load threshold, the load warning result and the operating condition prediction value and the energy consumption prediction value are determined and output, and the operating condition data with time series characteristics can be used to perform joint prediction through multiple series-connected models, and the prediction efficiency of multiple prediction data is improved by reducing the redundancy in the structures of multiple models, and the effective use of time series characteristics and the joint training of multiple series-connected models can improve the accuracy of the prediction value. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0040] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0041] Figure 1 A schematic diagram of a process flow of a multi-data joint prediction method provided by an embodiment of the present application is shown;
[0042] Figure 2 The schematic diagram of the structure of the model series connection provided by the embodiment of the present application is shown;
[0043] Figure 3 A schematic diagram of a multi-data joint prediction method provided by another embodiment of the present application is shown;
[0044] Figure 4 The detailed flow chart of step S13 provided in the embodiment of the present application is schematically shown;
[0045] Figure 5 The detailed flow chart of step S41 provided in the embodiment of the present application is schematically shown;
[0046] Figure 6 The schematic diagram shows a structure of a multi-data joint prediction device provided by an embodiment of the present application;
[0047] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown schematically. DETAILED DESCRIPTION
[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0049] See also Figure 1 The multi-data joint prediction method of the present application comprises the following steps:
[0050] S11, inputting historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model;
[0051] In practical applications, historical operating condition data with time series characteristics may be operating condition data sorted in the order of the time when the operating condition data was collected. The operating condition data may be one type of data or multiple types of data. When the operating condition data is one type of data, it may be any one of the temperature, voltage, power consumption, etc. in the environment and / or equipment operation at different times. When the operating condition data is multiple types of data, it may be at least two of the temperature, voltage, power consumption, etc. in the environment and / or equipment operation at different times. The operating condition prediction data may be operating condition data at a future time relative to the time when the historical operating condition data was collected. For example, the time when the historical operating condition data was collected was 8 a.m. every day in the past half month, and the time corresponding to the operating condition prediction data is 8 a.m. every day today, tomorrow, and any one or more days in the future.
[0052] S12, inputting the historical operating condition data and the operating condition prediction data with time series characteristics into a pre-trained energy consumption prediction model to obtain an energy consumption prediction value;
[0053] In practical applications, the relationship between the weight value of the energy consumption prediction model and the time corresponding to the operating condition data is: the earlier the time, the smaller the weight value of the operating condition data matching at that time, so that the newer operating condition data has a greater impact on the energy consumption prediction value.
[0054] S13, inputting historical operating condition data, operating condition prediction data and energy consumption prediction value with time series characteristics into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, such as Figure 2 As shown, joint training is performed during the model training process, and joint prediction is performed during the prediction process;
[0055] In practical applications, the relationship between the weight value of the load warning model and the time corresponding to the operating condition data is: the earlier the time, the smaller the weight value of the operating condition data matching at that time, so that the newer operating condition data has a greater impact on the load prediction value.
[0056] In practical applications, the load early warning model can be replaced by a fault prediction model, or the fault prediction model can be connected in parallel with the load early warning model.
[0057] When the equipment operates abnormally, fault information (including current operating condition data and fault type) is collected, including operating condition data when the fault occurs and a period of time before the fault occurs. The fault prediction model is trained based on this data to use the fault prediction model to timely perceive hidden dangers and possible equipment failure signals and issue alarms.
[0058] S14, determining and outputting a load warning result and the operating condition prediction value and the energy consumption prediction value according to a size relationship between the load prediction value and a preset load threshold.
[0059] In actual applications, the load prediction value is compared with the preset load threshold. When the load prediction value is greater than the preset load threshold, the load warning result is abnormal.
[0060] See also Figure 3 , the method further comprises:
[0061] S31, real-time monitoring of whether the increment between the data volume of the current working condition data, energy consumption value and load warning result and the data volume used in the last model training reaches the preset increment threshold:
[0062] In actual applications, when the new operating condition data, energy consumption values and load warning results accumulate to a certain amount compared with the data volume of the operating condition data, energy consumption values and load warning results used in the last model training, it is considered that the increment between the data volume of the current operating condition data, energy consumption values and load warning results and the data volume used in the last model training reaches the preset increment threshold.
[0063] If yes, execute step S32;
[0064] If not, execute step S33;
[0065] S32, using the incremental operating condition data, energy consumption value and load warning result to jointly update the operating condition prediction model, energy consumption prediction model and load warning model connected in series in sequence;
[0066] In practical applications, the joint update of the operating condition prediction model, energy consumption prediction model and load early warning model connected in series in sequence includes:
[0067] The known incremental historical operating condition data with time series characteristics is used as input, and the target operating condition data is used as output to train the operating condition prediction model to obtain a trained operating condition prediction model;
[0068] Inputting the known incremental historical operating condition data with time series characteristics into the trained operating condition prediction model to obtain the output value of the operating condition prediction model;
[0069] Taking the known incremental historical operating condition data with time series characteristics and the output value of the trained operating condition prediction model as input values, taking the known energy consumption value as output value, training the energy consumption prediction model, and obtaining the trained energy consumption prediction model;
[0070] The known incremental historical operating condition data with time series characteristics, the output value of the trained operating condition prediction model and the output value of the trained energy consumption prediction model are used as input values, and the known load prediction value is used as the output value to train the load warning model to obtain a trained load warning model.
[0071] S33, no response.
[0072] The multi-data prediction method of this embodiment can train the prediction model based on historical operating condition data, and perform multi-data prediction and multi-model update synchronously with new data on the equipment operating status, with high prediction efficiency and high accuracy of the prediction value.
[0073] See also Figure 4 In step S13, the joint training is performed during the model training process, including:
[0074] S41, using known historical operating condition data with time series characteristics as input and target operating condition data as output to train the operating condition prediction model to obtain a trained operating condition prediction model.
[0075] In practical applications, the operating condition prediction model is selected from a recurrent neural network, a long short-term memory network, and a threshold recurrent unit network according to historical operating condition data with time series characteristics.
[0076] In practical applications, the relationship between the initial value of the weight of the operating condition prediction model and the time corresponding to the operating condition data is: the earlier the time, the smaller the initial value of the weight matching the operating condition data at that time, so that the newer operating condition data has a greater impact on the operating condition prediction value.
[0077] S42, inputting known historical operating condition data with time series characteristics into the trained operating condition prediction model to obtain an output value of the operating condition prediction model.
[0078] S43, taking known historical operating condition data with time series characteristics and the output value of the trained operating condition prediction model as input values, taking known energy consumption values as output values, training the energy consumption prediction model, and obtaining a trained energy consumption prediction model.
[0079] In practical applications, the relationship between the initial value of the weight of the energy consumption prediction model and the time corresponding to the operating condition data is: the earlier the time, the smaller the initial value of the weight matching the operating condition data at that time, so that the newer operating condition data has a greater impact on the energy consumption prediction value.
[0080] In practical applications, the energy consumption prediction model is a multi-layer perceptron network or a convolutional neural network.
[0081] S44, taking known historical operating condition data with time series characteristics, the output value of the trained operating condition prediction model and the output value of the trained energy consumption prediction model as input values, and taking the known load prediction value as the output value, trains the load warning model to obtain a trained load warning model.
[0082] In practical applications, the relationship between the initial value of the weight of the load warning model and the time corresponding to the operating condition data is: the earlier the time, the smaller the initial value of the weight matching the operating condition data at that time, so that the newer operating condition data has a greater impact on the load prediction value.
[0083] In practical applications, the load warning model is a convolutional neural network, in which the initial weight values corresponding to the historical operating condition data with time series characteristics in the operating condition prediction model, energy consumption prediction model and load warning model present attenuation characteristics from back to front according to the time of the historical data.
[0084] See also Figure 5 In step S41, the operating condition prediction model is selected from a recurrent neural network, a long short-term memory network and a threshold recurrent unit network according to historical operating condition data with time series characteristics, including:
[0085] S51, during the joint training of the operating condition prediction model, the energy consumption prediction model and the load warning model connected in series in sequence, determining whether the time interval of the historical operating condition data with time series characteristics is less than a preset time threshold:
[0086] In practical applications, the time interval of the historical operating condition data with timing characteristics includes at least the difference between the moments corresponding to two adjacent historical operating condition data and the difference between the maximum and minimum moments corresponding to the historical operating condition data. When the difference between the moments corresponding to two adjacent historical operating condition data is less than a first preset time length threshold, and the difference between the maximum and minimum moments corresponding to the historical operating condition data is less than the first preset time length threshold, it is considered that the time interval of the historical operating condition data with timing characteristics is less than the preset time length threshold.
[0087] If yes, execute step S52;
[0088] If not, execute step S53;
[0089] S52, selecting a recurrent neural network as a working condition prediction model;
[0090] S53, select long short-term memory network as the working condition prediction model;
[0091] S54, determining whether the amount of historical operating condition data with time series characteristics is less than a preset data amount threshold:
[0092] If yes, execute step S55;
[0093] If not, return to step S53;
[0094] S55, selecting a threshold cycle unit network as an operating condition prediction model.
[0095] In practical applications, when the amount of historical operating condition data with time series characteristics is less than the preset data amount threshold and the time interval is less than the preset time length threshold, the bootstrap resampling method is used to resample the historical operating condition data with time series characteristics, so as to train the selected recurrent neural network based on the resampled historical operating condition data with time series characteristics.
[0096] The multi-data joint prediction method of this embodiment connects three models in series, and three independent tasks are trained separately. At the same time, the input of the previous model can be used as the output of the next model, completing a single construction of the entire complete system. It can output multiple prediction data in one prediction process, and the prediction efficiency is high. At the same time, operating condition data with time series characteristics are used in both model prediction and model training, and the operating condition prediction model adopts a time series neural network model, which can effectively utilize the time relationship information between the operating condition data at different times, so that the prediction value output by the model is more accurate.
[0097] The multi-data joint prediction method of this embodiment can automatically update the model according to the dynamically updated operating data, energy consumption values and load values, and synchronously predict the operating data, energy consumption values and load values at future times, so that the predicted value output by the model is more accurate and the prediction speed is faster.
[0098] The long short-term memory network in the multi-data joint prediction method of this embodiment can better capture long-distance and long time interval information than the recurrent neural network. Different models can be selected according to the length of the time series corresponding to the operating condition data to achieve the flexibility of the model to adapt to the data characteristics.
[0099] The multi-data joint prediction method of this embodiment adopts a time series neural network model (long short-term memory network or recurrent neural network) to effectively utilize the time series information of each sensor signal and can more accurately complete the prediction of operating point / energy consumption.
[0100] The multi-data joint prediction method of this embodiment utilizes the equipment operation status data within a period of time and is trained based on a neural network. It can predict the equipment operation status curve within a period of time in the future, and at the same time calculate and predict the energy consumption and efficiency based on the equipment status, thereby improving the prediction efficiency and the accuracy of the prediction value.
[0101] like Figure 6 As shown, an embodiment of the present application provides a multi-data joint prediction device, which includes: an operating condition prediction module 11, an energy consumption prediction module 12, a load prediction module 13 and an output module 14.
[0102] In this embodiment, the operating condition prediction module 11 is used to input historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model;
[0103] In this embodiment, the energy consumption prediction module 12 is used to input the historical operating condition data and operating condition prediction data with time series characteristics into a pre-trained energy consumption prediction model to obtain an energy consumption prediction value;
[0104] In this embodiment, the load prediction module 13 is used to input the historical working condition data, working condition prediction data and energy consumption prediction value with time series characteristics into a pre-trained load warning model to obtain a load prediction value, wherein the working condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed in the model training process, and joint prediction is performed in the prediction process;
[0105] In this embodiment, the output module 14 is used to determine and output the load warning result and the operating condition prediction value and the energy consumption prediction value according to the size relationship between the load prediction value and the preset load threshold.
[0106] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.
[0107] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.
[0108] Based on the same inventive concept, Figure 7 As shown, an embodiment of the present application provides an electronic device, including a processor 1110, a communication interface 1120, a memory 1130 and a communication bus 1140, wherein the processor 1110, the communication interface 1120, and the memory 1130 communicate with each other through the communication bus 1140;
[0109] Memory 1130, used for storing computer programs;
[0110] The processor 1110 is used to implement the following multi-data joint prediction method when executing the program stored in the memory 1130:
[0111] Inputting historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model;
[0112] Input the historical operating condition data and operating condition prediction data with time series characteristics into the pre-trained energy consumption prediction model to obtain the energy consumption prediction value;
[0113] Inputting historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process;
[0114] According to the size relationship between the load prediction value and the preset load threshold, the load warning result and the operating condition prediction value and the energy consumption prediction value are determined and output.
[0115] The communication bus 1140 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus 1140 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
[0116] The communication interface 1120 is used for communication between the above electronic device and other devices.
[0117] The memory 1130 may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory 1130 may also be at least one storage device located away from the processor 1110.
[0118] The above-mentioned processor 1110 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components.
[0119] An embodiment of the present application provides a computer-readable storage medium, which stores one or more programs. The one or more programs can be executed by one or more processors to implement the multi-data joint prediction method in any possible implementation method described above.
[0120] Alternatively, the storage medium may be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
[0121] Based on the same inventive concept, an embodiment of the present application further provides a computer-readable storage medium, including a computer program, which, when executed by a processor, implements the multi-data joint prediction method in any possible implementation manner described above.
[0122] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loading and executing computer program instructions on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server, a data center, etc. that contains one or more available media integrated. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.
[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A multi-data joint prediction method, characterized in that: The following steps are involved: Inputting historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model; Input the historical operating condition data and operating condition prediction data with time series characteristics into the pre-trained energy consumption prediction model to obtain the energy consumption prediction value; Inputting historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process; According to the magnitude relationship between the load prediction value and the preset load threshold, the load warning result and the operating condition prediction value and the energy consumption prediction value are determined and output.
2. The method according to claim 1, characterized in that The method further comprises: Real-time monitoring of the data volume of current operating condition data, energy consumption value, and load warning results and whether the increment between the data volume used in the last model training reaches the preset increment threshold: When the increment between the amount of data of the current operating condition data, energy consumption value and load warning results and the amount of data used in the last model training reaches the preset increment threshold, the incremental operating condition data, energy consumption value and load warning results are used to jointly update the operating condition prediction model, energy consumption prediction model and load warning model connected in series.
3. The method according to claim 1, characterized in that The joint training is performed during the model training process, including: Using known historical operating condition data with time series characteristics as input and target operating condition data as output to train the operating condition prediction model, a trained operating condition prediction model is obtained; Input known historical operating condition data with time series characteristics into the trained operating condition prediction model to obtain the output value of the operating condition prediction model; Taking known historical operating condition data with time series characteristics and the output value of the trained operating condition prediction model as input values, taking known energy consumption values as output values, training the energy consumption prediction model, and obtaining a trained energy consumption prediction model; The known historical operating condition data with time series characteristics, the output value of the trained operating condition prediction model and the output value of the trained energy consumption prediction model are used as input values, and the known load prediction value is used as the output value to train the load warning model to obtain a trained load warning model.
4. The method according to claim 3, characterized in that The operating condition prediction model is selected from a recurrent neural network, a long short-term memory network, and a threshold recurrent unit network according to historical operating condition data with time series characteristics; the energy consumption prediction model is a multi-layer perceptron network or a convolutional neural network; the load warning model is a convolutional neural network, wherein the initial weight values corresponding to the historical operating condition data with time series characteristics in the operating condition prediction model, the energy consumption prediction model, and the load warning model present a decay characteristic from back to front according to the time of the historical data.
5. The method according to claim 4, characterized in that The operating condition prediction model is selected from a recurrent neural network, a long short-term memory network, and a threshold recurrent unit network according to historical operating condition data with time series characteristics, and includes: During the joint training of the operating condition prediction model, energy consumption prediction model and load warning model connected in series in sequence, when the time interval of the historical operating condition data with time series characteristics is less than a preset time threshold, selecting a recurrent neural network as the operating condition prediction model; When the time interval of historical operating condition data with time series characteristics is greater than or equal to the preset time threshold, the long short-term memory network is selected as the operating condition prediction model; When the amount of historical operating condition data with time series characteristics is less than a preset data amount threshold and the time interval is greater than or equal to a preset time length threshold, the threshold cycle unit network is selected as the operating condition prediction model.
6. The method according to claim 5, characterized in that The method further comprises: When the amount of historical operating condition data with time series characteristics is less than a preset data amount threshold and the time interval is less than a preset time length threshold, the bootstrap resampling method is used to resample the historical operating condition data with time series characteristics, so as to train the selected recurrent neural network based on the resampled historical operating condition data with time series characteristics.
7. The method according to claim 1, characterized in that The method further comprises: The historical operating condition data, operating condition prediction data and energy consumption prediction values with time series characteristics are input into a pre-trained fault prediction model, and a fault warning result is output, wherein the fault prediction model is connected in series with the energy consumption prediction model so that the input value of the fault prediction model includes the output value of the energy consumption prediction model.
8. A multi-data joint prediction device, characterized in that: include: An operating condition prediction module, which is used to input historical operating condition data with time series characteristics into a pre-trained operating condition prediction model to obtain operating condition prediction data predicted by the operating condition prediction model; An energy consumption prediction module is used to input historical operating condition data and operating condition prediction data with time series characteristics into a pre-trained energy consumption prediction model to obtain energy consumption prediction values; A load prediction module, which is used to input historical operating condition data with time series characteristics, operating condition prediction data and energy consumption prediction values into a pre-trained load warning model to obtain a load prediction value, wherein the operating condition prediction model, the energy consumption prediction model and the load warning model are connected in series in sequence, and joint training is performed during the model training process, and joint prediction is performed during the prediction process; The output module is used to determine and output the load warning result and the working condition prediction value and energy consumption prediction value according to the size relationship between the load prediction value and the preset load threshold.
9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor is used to implement the steps of the multi-data joint prediction method described in any one of claims 1 to 7 when executing a program stored in a memory.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the multiple data joint prediction method according to any one of claims 1 to 7 are implemented.