Non-intrusive load monitoring method, device, equipment, medium and product

Abstract

The invention relates to a non-intrusive load monitoring method, device and equipment, a medium and a product. The method comprises the following steps: carrying out coding processing on total load time sequence data collected by non-intrusive load monitoring equipment to obtain coding feature data; scaling the coded feature data to obtain scaled feature data; performing sampling prediction processing on the zoom feature data through a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time sequence data; and performing load value estimation processing based on the equipment load data sequence, and determining load value prediction results of the multiple pieces of equipment acquired by the non-intrusive load monitoring equipment. According to the invention, load data sampling and load value estimation processing are carried out based on the fine-tuned large language model, the load data of each device is effectively speculated, the accuracy of load identification in the load monitoring process is effectively improved, and the load monitoring effect is ensured.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a non-invasive load monitoring method, device, equipment, medium and product. Background Art

[0002] Nowadays, energy demand is growing rapidly, and energy efficiency has become an important issue. In order to achieve accurate monitoring of electricity consumption behavior, non-intrusive load monitoring (NILM) methods have attracted people's attention and achieved rapid development. Non-intrusive load monitoring technology only requires a single sensor to be installed at the power entrance of each household. It has low deployment cost and small installation complexity, and has broad application prospects.

[0003] However, current non-intrusive load monitoring methods are generally based on deep learning models, with low prediction accuracy and unable to guarantee the load identification effect of the load monitoring process. Summary of the invention

[0004] Based on this, it is necessary to provide a non-invasive load monitoring method, device, equipment, medium and product that can improve the accuracy of non-invasive load monitoring in response to the above technical problems.

[0005] In a first aspect, the present application provides a non-intrusive load monitoring method, comprising:

[0006] Encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain coded feature data;

[0007] Scaling the encoded feature data to obtain scaled feature data;

[0008] The scaled feature data is sampled and predicted by a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model using historical load data;

[0009] Load value estimation processing is performed based on the equipment load data sequence to determine load value prediction results of multiple equipment collected by the non-intrusive load monitoring device.

[0010] In one embodiment, the encoding process of the total load data collected by the non-intrusive load monitoring device to obtain the encoded characteristic data includes:

[0011] Obtaining time series data corresponding to total load data collected by non-intrusive load monitoring equipment;

[0012] Encoding the time series data to obtain an encoded character string;

[0013] The encoding character string is adjusted based on preset data accuracy requirements to obtain encoding feature data.

[0014] In one embodiment, the performing load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring device includes:

[0015] Determining the median of each time step in the equipment load data series;

[0016] The median is used as a point estimate of the time step to determine the load value prediction results of the multiple devices collected by the non-intrusive load monitoring device.

[0017] In one embodiment, the performing load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring device includes:

[0018] Converting the equipment load data sequence into a load value probability density function;

[0019] A load value estimation process is performed based on the load value probability density function to determine load value prediction results of multiple devices collected by the non-intrusive load monitoring device.

[0020] In one embodiment, converting the equipment load data sequence into a load value probability density function comprises:

[0021] Convert each time step output in the equipment load data sequence into a digital sequence;

[0022] Determine the probability density representation corresponding to the digital sequence, and obtain the load value probability density function corresponding to the equipment load data sequence;

[0023] The performing load value estimation processing based on the load value probability density function to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device includes:

[0024] Integrating the load value probability density function corresponding to each device to determine the optimal estimate of the load data of the device;

[0025] The optimal estimation of the load data of each device is summarized to obtain the load value prediction results of the multiple devices collected by the non-intrusive load monitoring device.

[0026] In one embodiment, the method further comprises:

[0027] Obtaining a historical load data set including total load data at the bus and load data of each device;

[0028] Performing encoding and scaling processing on the load data in the historical load data set to obtain load encoding data of the historical load data set;

[0029] constructing a fine-tuning training set based on the load-encoded data;

[0030] The pre-trained large language model is fine-tuned using the fine-tuning training set to obtain a preset large language model.

[0031] In a second aspect, the present application also provides a non-intrusive load monitoring device, comprising:

[0032] A data encoding module is used to encode the total load time series data collected by the non-intrusive load monitoring equipment to obtain encoding feature data;

[0033] A data scaling module, used for scaling the encoded feature data to obtain scaled feature data;

[0034] A model processing module, used for performing sampling prediction processing on the scaled feature data through a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model through historical load data;

[0035] The load prediction module is used to perform load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment.

[0036] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0037] Encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain coded feature data;

[0038] Scaling the encoded feature data to obtain scaled feature data;

[0039] The scaled feature data is sampled and predicted by a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model using historical load data;

[0040] Load value estimation processing is performed based on the equipment load data sequence to determine load value prediction results of multiple equipment collected by the non-intrusive load monitoring device.

[0041] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0042] Encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain coded feature data;

[0043] Scaling the encoded feature data to obtain scaled feature data;

[0044] The scaled feature data is sampled and predicted by a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model using historical load data;

[0045] Load value estimation processing is performed based on the equipment load data sequence to determine load value prediction results of multiple equipment collected by the non-intrusive load monitoring device.

[0046] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0047] Encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain coded feature data;

[0048] Scaling the encoded feature data to obtain scaled feature data;

[0049] The scaled feature data is sampled and predicted by a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model using historical load data;

[0050] Load value estimation processing is performed based on the equipment load data sequence to determine load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment.

[0051] The above-mentioned non-intrusive load monitoring method, device, computer equipment, computer-readable storage medium and computer program product obtain encoded feature data by encoding the total load time series data collected by the non-intrusive load monitoring device; scale the encoded feature data to obtain scaled feature data; sample and predict the scaled feature data through a preset large language model to obtain multiple equipment load data sequences corresponding to the total load time series data, and the preset large language model is obtained by fine-tuning the pre-trained large language model through historical load data; load value estimation processing is performed based on the equipment load data sequence to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device. This application realizes the efficient adaptability of the model through feature encoding and data scaling processing, so that the preset large language model can be used for load monitoring processing, thereby effectively combining the generalization ability of the pre-trained large language model and the load recognition ability of the fine-tuned model. Based on the fine-tuned large language model, the load data of each device is effectively inferred, which effectively improves the accuracy of load recognition in the load monitoring process and ensures the load monitoring effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0053] FIG1 is a diagram of an application environment of a non-intrusive load monitoring method in one embodiment;

[0054] FIG2 is a schematic diagram of a flow chart of a non-intrusive load monitoring method in one embodiment;

[0055] FIG3 is a schematic diagram of a flow chart of a feature encoding process in one embodiment;

[0056] FIG4 is a schematic diagram of a process flow of a model fine-tuning training process in one embodiment;

[0057] FIG5 is a schematic flow chart of a non-intrusive load monitoring method in another embodiment;

[0058] FIG6 is a schematic diagram of the load decomposition effect of the kiln head transformer in one embodiment;

[0059] FIG7 is a schematic diagram of the load decomposition effect of the kiln exhaust fan in one embodiment;

[0060] FIG8 is a schematic diagram of the load decomposition effect of the main motor of a coal mill in one embodiment;

[0061] FIG9 is a block diagram of a non-intrusive load monitoring device according to an embodiment;

[0062] FIG. 10 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0064] The non-intrusive load monitoring method provided by the embodiment of the present application can be applied to the application environment shown in Figure 1. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. When the staff responsible for load monitoring wants to analyze the collected load monitoring data, they can submit a corresponding request to the server 104 through the terminal 102, and the server 104 performs corresponding monitoring data analysis and processing according to the request. The server 104 first obtains the total load time series data collected by the non-intrusive load monitoring device; encodes the total load time series data collected by the non-intrusive load monitoring device to obtain the encoded feature data; scales the encoded feature data to obtain the scaled feature data; samples and predicts the scaled feature data through a preset large language model to obtain multiple device load data sequences corresponding to the total load time series data, and the preset large language model is obtained by fine-tuning the pre-trained large language model through historical load data; load value estimation processing is performed based on the device load data sequence to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device. The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. Head-mounted devices may be virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc. The server 104 may be an independent physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0065] In an exemplary embodiment, as shown in FIG2 , a non-intrusive load monitoring method is provided, which is described by taking the method applied to the server 104 in FIG1 as an example, and includes the following steps 201 to 207. Among them:

[0066] Step 201 , encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain encoding feature data.

[0067] Step 203: scaling the encoded feature data to obtain scaled feature data.

[0068] Among them, non-intrusive load monitoring equipment refers to equipment that collects load monitoring related data in a non-intrusive manner. Non-intrusive load monitoring equipment monitors and analyzes the voltage, current and other data at the load entrance of the power user to realize functions such as user power equipment category identification and itemized power energy calculation. By analyzing and processing the total load meter data, it is possible to identify each power equipment and its working status. The total load time series data refers to a sequence of total load data at different times, which can be expressed as ,in Indicates time The encoding process is to facilitate the processing of the large language model, and the total load time series data is encoded and processed to obtain the encoded feature data composed of strings. As for the scaling process, in order to enable the subsequent large language model to have a good numerical range processing capability when processing time series data, it is necessary to scale the data to avoid the data range being too large or too small.

[0069] Exemplarily, for the process of load monitoring, when the staff responsible for load monitoring wants to analyze the collected load monitoring data, the corresponding load monitoring processing request can be submitted to the server 104 through the terminal 102, and the data to be analyzed can be specified in the request, and then the server 104 performs data search processing according to the request to obtain the total load time series data collected by the non-intrusive load monitoring device as the basis for data analysis. For the convenience of subsequent processing, the total load time series data collected can be encoded and converted into encoded feature data. Then the obtained encoded feature data is scaled to adjust the data range of these data. In one embodiment, the processing process for non-intrusive load monitoring can be automatically triggered by cycle, such as collecting and analyzing data by day or week. When a new cycle is reached, the server 104 can obtain the total load time series data collected by the non-intrusive load monitoring device in the previous cycle, so as to analyze and process the load monitoring of the previous collection cycle. As for the scaling processing of features, in order to enable the large language model to have good numerical range processing capabilities when processing time series data, it is necessary to scale these data to avoid the data range being too large or too small. Set zoom factor and offset , for the scaling process, first calculate the minimum value of the time series data and maximum value , and offset the data using the following formula:

[0070]

[0071] in, is the offset, The adjustment coefficient is used to control the magnitude of data offset to ensure that the data does not exceed a reasonable range during the scaling process.

[0072] After the offset processing is completed, the scaling factor is used Scale the offset time series data. Suppose the adjusted data is , then the scaled data It is expressed as:

[0073]

[0074] in, Adjusted time series data No. Through this scaling process, the subsequent large language model can better handle the magnitude of changes in time series data.

[0075] Step 205, sampling and predicting the scaled feature data is processed by a preset large language model to obtain multiple equipment load data sequences corresponding to the total load time series data. The preset large language model is obtained by fine-tuning the pre-trained large language model through historical load data.

[0076] Among them, the Large Language Model (LLM) is a language model composed of a neural network with many parameters (usually billions of weights or more), which is trained on a large amount of unlabeled text using self-supervised learning or semi-supervised learning. Generally, a large amount of text data can be collected to complete the pre-training of the large language model, and then the large language model can be fine-tuned in combination with specific tasks. Therefore, for the large language model used in the present application, the pre-trained large language model can be fine-tuned and trained using historical load data. The large language model can decompose the load characteristics of each device from the total load data, and the separation method is sampling prediction, which generates multiple possible device load data sequences through multiple samplings, from which the load characteristics of the device can be inferred.

[0077] Exemplarily, the present application specifically uses a large language model to complete the load decomposition processing of the load monitoring process. Therefore, after obtaining the pre-processed scaled feature data, the scaled feature data can be input into the preset large language model. The preset large language model can infer the load data of each device based on the total load data. When the re-scaled scaled feature data is used as input, the preset large language model can output the corresponding multiple device load data sequences through multiple sampling processes. Assume that each time step The output is , then the sequence is generated It can be generated by the following conditional probability chain:

[0078]

[0079] in, is the predicted time step, which indicates the length of the model's prediction for future time steps. Multiple possible equipment load data sequences are generated through multiple samplings, from which the load characteristics of the equipment can be inferred.

[0080] Step 207 , performing load value estimation processing based on the equipment load data sequence, and determining load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment.

[0081] For example, after obtaining the predicted equipment load data sequence, the load value estimation processing can be performed on these equipment load data sequences to obtain the load data estimate. For the estimation process of the specific load value, the estimation can be completed directly in the form of point estimation, such as using the median as the point estimate, or by calculating the quantile of each time step to construct a prediction confidence interval, and determine the load value prediction results of multiple devices collected by the non-invasive load monitoring equipment. In addition, if you want to achieve more accurate load estimation, you can further improve the accuracy of the prediction through continuous probability density modeling, that is, through continuous probability density modeling to construct the probability density function of each equipment load data sequence, to achieve complex multi-modal modeling of the load data, and thus improve the accuracy and reliability of the prediction.

[0082] The above-mentioned non-intrusive load monitoring method obtains coded feature data by encoding the total load time series data collected by the non-intrusive load monitoring device; scaling the coded feature data to obtain scaled feature data; sampling and predicting the scaled feature data through a preset large language model to obtain multiple equipment load data sequences corresponding to the total load time series data, and the preset large language model is obtained by fine-tuning the pre-trained large language model through historical load data; load value estimation processing is performed based on the equipment load data sequence to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device. This application realizes the efficient adaptability of the model through feature encoding and data scaling processing, so that the preset large language model can be used for load monitoring processing, thereby effectively combining the generalization ability of the pre-trained large language model and the load recognition ability of the fine-tuned model. Based on the fine-tuned large language model, the load data of each device is effectively inferred, which effectively improves the accuracy of load recognition in the load monitoring process and ensures the load monitoring effect.

[0083] In an exemplary embodiment, as shown in FIG3 , step 203 includes:

[0084] Step 302: Acquire time series data corresponding to total load data collected by non-intrusive load monitoring equipment.

[0085] Step 304: Encode the time series data to obtain an encoded character string.

[0086] Step 306: Adjust the encoding string based on the preset data accuracy requirement to obtain encoding feature data.

[0087] For example, in the feature coding process, after obtaining the time series data corresponding to the total load data collected by the non-intrusive load monitoring device, in order to enable the large language model to correctly process these time series data, it is necessary to tokenize them. For the collected time series data, it can be expressed as ,in Indicates time The encoding process is to convert the time series data into a string. Convert to string form Assume that the data accuracy is decimal places, then It can be expressed as:

[0088]

[0089] And will When represented as a string, spaces can be inserted between each number so that the large language model can tokenize them independently. For example, if , then it is expressed as . At the same time, for the entire time series All data points in , using commas "," to separate the data values ​​of each time step to form the final encoded string :

[0090]

[0091] During the data conversion process, the encoding string can also be adjusted based on the preset data precision requirements to obtain the encoding feature data. For example, for the value after the decimal point, the decimal point can be discarded according to the preset precision to reduce the length of the string. If the precision is set to two decimal places, Can be encoded as a string , so that the monitoring process can focus on the most significant features and effectively improve the model recognition accuracy. In this embodiment, the encoding and string adjustment of time series data are used to construct encoding feature data that is more suitable for the large language model, thereby ensuring the accuracy and efficiency of the subsequent load sampling process and improving the accuracy of load monitoring.

[0092] In an exemplary embodiment, step 207 includes: determining the median of each time step in the equipment load data sequence; using the median as a point estimate of the time step to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device.

[0093] For example, for the load forecasting process, after obtaining the equipment load data sequence, prediction can be made directly based on the equipment load data sequence, that is, the median of each time step in the equipment load data sequence is determined, and this median is used as the point estimate of the current time step. After processing each time step, the load value prediction results of multiple devices collected by the non-intrusive load monitoring device can be obtained. Specifically, by generating multiple equipment load data sequences By conducting statistical analysis, the corresponding load value prediction results can be obtained.

[0094] The median of each time step can be calculated As a point estimate, that is:

[0095]

[0096] in, is the sampling number, For the final device The median of the load forecast value. In another embodiment, the prediction confidence interval can also be constructed by calculating the quantile of each time step, so as to perform point estimation to determine the load value prediction result. In this embodiment, the point estimation is performed by the median of the time step to obtain the load value prediction result, which can effectively improve the load value prediction efficiency of the load monitoring process and ensure the processing speed.

[0097] In an exemplary embodiment, step 207 includes: converting the equipment load data sequence into a load value probability density function; performing load value estimation processing based on the load value probability density function to determine load value prediction results of multiple devices collected by non-intrusive load monitoring equipment.

[0098] Among them, for the probability density function, in mathematics, the probability density function of a continuous random variable (which can be referred to as the density function when there is no confusion) is a function that describes the probability of the output value of this random variable near a certain value point. The probability that the value of the random variable falls within a certain area is the integral of the probability density function over this area. When the probability density function exists, the cumulative distribution function is the integral of the probability density function.

[0099] Exemplarily, in order to more accurately identify and predict the load characteristics of different devices in non-invasive load monitoring, the present application can convert the discrete large language model output into a continuous probability density function to form a refined modeling of the load data to further improve the accuracy of load prediction. First, it is necessary to convert the device load data sequence into a load value probability density function. For this process, in one embodiment, each time step output in the device load data sequence can be converted into a digital sequence; the probability density representation corresponding to the digital sequence is determined to obtain the load value probability density function corresponding to the device load data sequence. For example, for the device load data sequence , assuming that each data point Can be expressed as a cardinality of Digit sequence , and each digit The probability of generating ,but The probability of generating is:

[0100]

[0101] in, is the first Bit, Indicates that before it is known Digits and input sequences Under the conditions of Digit The probability of generation. By Convert to a continuous probability density representation, using a uniform distribution in the digital interval to convert the discrete generation results Convert to a continuous load value probability density function :

[0102]

[0103] in, For the Uniform distribution over a number interval, is the discrete generation probability in the interval, which means Then, the load value estimation process can be performed based on the load value probability density function to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device. For the load value prediction process, the load value probability density function corresponding to each device can be integrated to determine the optimal load data estimate of the device. , which can be used to perform complex multi-modal modeling on load data, thereby improving the accuracy and reliability of prediction. The final prediction value can be calculated by the following formula:

[0104]

[0105] in, Indicates future equipment The optimal estimate of the load data, is the load value The probability density function of can be integrated to obtain the expected value, i.e., the final predicted value. Finally, the optimal estimate of the load data of each device is summarized to obtain the load value prediction results of multiple devices collected by the non-intrusive load monitoring device. In this embodiment, by modeling the continuous probability density of the equipment load data sequence, the load data of each device can be effectively inferred, which effectively improves the accuracy of load identification in the load monitoring process and ensures the load monitoring effect.

[0106] In the first embodiment, as shown in FIG4 , the method further includes:

[0107] Step 401 : Acquire a historical load data set including total load data at the bus and load data of each device.

[0108] Step 403 , encoding and scaling the load data in the historical load data set to obtain load encoding data of the historical load data set.

[0109] Step 405: construct a fine-tuning training set based on the load-coded data.

[0110] Step 407 , fine-tune the large language model obtained by pre-training by fine-tuning the training set to obtain a preset large language model.

[0111] Among them, the historical load data set of the total load data at the bus and the load data of each device can specifically be the actual cement plant kiln head and coal mill collection data set. The collected cement plant kiln head and coal mill data set contains the total load data at the bus and the load data of each device. Through these historical load data, the fine-tuning training processing of the large language model can be completed, thereby ensuring the efficiency and accuracy of the model in processing load data.

[0112] For example, before the load recognition process is performed through the large language model, the large language model needs to be fine-tuned. The fine-tuning process requires the collection of sample data for model training. Here, a historical load data set including the total load data at the bus and the load data of each device can be selected, and sample data can be constructed based on the historical load data set. The total load data at the bus and the load data of each device, the two time series data can be expressed as and ,in Indicates time The total load value, Indicates the device At the moment Before fine-tuning the training, these time series data also need to be encoded and scaled to be consistent with the input data of the model. The load data in the historical load data set needs to be encoded and scaled. Finally, the encoded string is obtained. and By converting the total load data Load data of each device Combined to form a fine-tuning training set. Then fine-tune the large language model, which can be done by minimizing the following loss function Fine-tune the model:

[0113]

[0114] in, For the model based on the total load data Predicted Equipment The load value, are the parameters of the model. During the training process, the model optimizes the parameters The ability to extract equipment load features from the total load data is learned until a usable large language model is obtained. In this embodiment, the large language model is fine-tuned by collecting historical load data sets to construct training samples, thereby ensuring the load recognition accuracy of the preset large language model.

[0115] In one embodiment, the overall process of the non-invasive load monitoring method of the present application is shown in Figure 5, and the specific steps include: obtaining the time series data corresponding to the total load data collected by the non-invasive load monitoring equipment; encoding the time series data to obtain a coded string; adjusting the coded string based on the preset data accuracy requirements, deleting redundant information, and obtaining coded feature data. Scaling the coded feature data to obtain scaled feature data. Obtain a historical load data set including the total load data at the bus and the load data of each device; encode and scale the load data in the historical load data set to obtain the load coded data of the kiln head and coal mill data set actually collected by the cement plant; construct a fine-tuning training set based on the load coded data; fine-tune the large language model obtained by pre-training through the fine-tuning training set to obtain a preset large language model. Sampling and predicting the scaled feature data through the preset large language model to obtain multiple equipment load data sequences corresponding to the total load time series data. Convert each time step output in the equipment load data sequence into a digital sequence; determine the probability density representation corresponding to the digital sequence to obtain the load value probability density function corresponding to the equipment load data sequence; integrate the load value probability density function corresponding to each device to determine the optimal estimate of the load data of the device; summarize the optimal estimate of the load data of each device to obtain the load value prediction results of multiple devices collected by non-intrusive load monitoring equipment.

[0116] This application is based on the fine-tuning of the kiln head and coal mill data sets actually collected from the cement plant. The model can learn the load characteristics of each device from the total load data, and realize the efficient adaptability of the model through encoding and data scaling processing. Based on the fine-tuned large language model, the load data is sampled and continuous probability density modeling is carried out to effectively infer the load data of each device, avoiding the limitations of the traditional load monitoring method based on deep learning models, and realizing high-precision non-invasive load monitoring. The effect of load decomposition for the kiln head and coal mill data sets can be specifically shown in Figures 6, 7, and 8, where Figure 6 is the load decomposition effect of the kiln head transformer, Figure 7 is the load decomposition effect of the kiln head exhaust fan, and Figure 8 is the load decomposition effect of the coal mill main motor. From the figure, it can be determined that the present application effectively realizes non-invasive load monitoring processing for the kiln head and coal mill data sets.

[0117] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0118] Based on the same inventive concept, the embodiment of the present application also provides a non-intrusive load monitoring device for implementing the non-intrusive load monitoring method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more non-intrusive load monitoring device embodiments provided below can refer to the limitations of the non-intrusive load monitoring method above, and will not be repeated here.

[0119] In an exemplary embodiment, as shown in FIG9 , a non-intrusive load monitoring device is provided, comprising:

[0120] The data encoding module 902 is used to encode the total load time series data collected by the non-intrusive load monitoring equipment to obtain encoded feature data.

[0121] The data scaling module 904 is used to scale the encoded feature data to obtain scaled feature data.

[0122] The model processing module 906 is used to perform sampling and prediction processing on the scaled feature data through a preset large language model to obtain multiple equipment load data sequences corresponding to the total load time series data. The preset large language model is obtained by fine-tuning the pre-trained large language model through historical load data.

[0123] The load prediction module 908 is used to perform load value estimation processing based on the equipment load data sequence, and determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment.

[0124] In one embodiment, the data encoding module 902 is specifically used to: obtain time series data corresponding to the total load data collected by the non-invasive load monitoring equipment; encode the time series data to obtain a coding string; adjust the coding string based on preset data accuracy requirements to obtain coding feature data.

[0125] In one embodiment, the load prediction module 908 is specifically used to: determine the median of each time step in the equipment load data sequence; use the median as a point estimate of the time step to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device.

[0126] In one embodiment, the load prediction module 908 is specifically used to: convert the equipment load data sequence into a load value probability density function; perform load value estimation processing based on the load value probability density function, and determine the load value prediction results of multiple devices collected by non-invasive load monitoring equipment.

[0127] In one embodiment, the load prediction module 908 is specifically used to: convert each time step output in the equipment load data sequence into a digital sequence; determine the probability density representation corresponding to the digital sequence to obtain the load value probability density function corresponding to the equipment load data sequence; integrate the load value probability density function corresponding to each device to determine the optimal estimate of the load data of the device; summarize the optimal estimate of the load data of each device to obtain the load value prediction results of multiple devices collected by the non-invasive load monitoring device.

[0128] In one embodiment, it also includes a model fine-tuning module, which is used to: obtain a historical load data set including total load data at the bus and load data of each device; encode and scale the load data in the historical load data set to obtain load encoding data of the historical load data set; construct a fine-tuning training set based on the load encoding data; and fine-tune the large language model obtained by pre-training through the fine-tuning training set to obtain a preset large language model.

[0129] Each module in the above non-intrusive load monitoring device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0130] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG10. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to non-intrusive load monitoring. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a non-intrusive load monitoring method is implemented.

[0131] Those skilled in the art will understand that the structure shown in FIG. 10 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0132] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0133] In one embodiment, a computer-readable storage medium is provided, storing a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0134] In one embodiment, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps in the above-mentioned method embodiments.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0136] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0138] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A non-intrusive load monitoring method, characterized in that: The method comprises: Encoding the total load time series data collected by the non-intrusive load monitoring equipment to obtain coded feature data; Scaling the encoded feature data to obtain scaled feature data; The scaled feature data is sampled and predicted by a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model using historical load data; Load value estimation processing is performed based on the equipment load data sequence to determine load value prediction results of multiple equipment collected by the non-intrusive load monitoring device.

2. The method according to claim 1, characterized in that The encoding process of the total load data collected by the non-intrusive load monitoring device to obtain the encoded characteristic data includes: Obtaining time series data corresponding to total load data collected by non-intrusive load monitoring equipment; Encoding the time series data to obtain an encoded character string; The encoding character string is adjusted based on preset data accuracy requirements to obtain encoding feature data.

3. The method according to claim 1, characterized in that The performing load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment includes: Determining the median of each time step in the equipment load data series; The median is used as a point estimate of the time step to determine the load value prediction results of the multiple devices collected by the non-intrusive load monitoring device.

4. The method according to claim 1, characterized in that The performing load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment includes: Converting the equipment load data sequence into a load value probability density function; A load value estimation process is performed based on the load value probability density function to determine load value prediction results of multiple devices collected by the non-intrusive load monitoring device.

5. The method according to claim 4, characterized in that The converting the equipment load data sequence into a load value probability density function comprises: Convert each time step output in the equipment load data sequence into a digital sequence; Determine the probability density representation corresponding to the digital sequence, and obtain the load value probability density function corresponding to the equipment load data sequence; The performing load value estimation processing based on the load value probability density function to determine the load value prediction results of multiple devices collected by the non-intrusive load monitoring device includes: Integrating the load value probability density function corresponding to each device to determine the optimal estimate of the load data of the device; The optimal estimation of the load data of each device is summarized to obtain the load value prediction results of the multiple devices collected by the non-intrusive load monitoring device.

6. The method according to any one of claims 1 to 5, characterized in that: The method further comprises: Obtaining a historical load data set including total load data at the bus and load data of each device; Performing encoding and scaling processing on the load data in the historical load data set to obtain load encoding data of the historical load data set; constructing a fine-tuning training set based on the load-encoded data; The pre-trained large language model is fine-tuned using the fine-tuning training set to obtain a preset large language model.

7. A non-intrusive load monitoring device, characterized in that: The device comprises: A data encoding module is used to encode the total load time series data collected by the non-intrusive load monitoring equipment to obtain encoding feature data; A data scaling module, used for scaling the encoded feature data to obtain scaled feature data; A model processing module, used for performing sampling prediction processing on the scaled feature data through a preset large language model to obtain a plurality of equipment load data sequences corresponding to the total load time series data, wherein the preset large language model is obtained by fine-tuning a pre-trained large language model through historical load data; The load prediction module is used to perform load value estimation processing based on the equipment load data sequence to determine the load value prediction results of multiple equipment collected by the non-intrusive load monitoring equipment.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. 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 method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images