CN117578487A - Non-invasive load monitoring method, device and computer equipment - Google Patents

Non-invasive load monitoring method, device and computer equipment Download PDF

Info

Publication number
CN117578487A
CN117578487A CN202311527771.1A CN202311527771A CN117578487A CN 117578487 A CN117578487 A CN 117578487A CN 202311527771 A CN202311527771 A CN 202311527771A CN 117578487 A CN117578487 A CN 117578487A
Authority
CN
China
Prior art keywords
neural network
power state
convolution neural
power
state
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202311527771.1A
Other languages
Chinese (zh)
Inventor
何耿生
曾金灿
姚尚衡
张舒涵
杨鑫和
李沛
梁梓杨
刘玺
黄宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guizhou Power Grid Co Ltd
Energy Development Research Institute of China Southern Power Grid Co Ltd
Original Assignee
Guizhou Power Grid Co Ltd
Energy Development Research Institute of China Southern Power Grid Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guizhou Power Grid Co Ltd, Energy Development Research Institute of China Southern Power Grid Co Ltd filed Critical Guizhou Power Grid Co Ltd
Priority to CN202311527771.1A priority Critical patent/CN117578487A/en
Publication of CN117578487A publication Critical patent/CN117578487A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
    • H02J13/00006—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment
    • H02J13/00016—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network characterised by information or instructions transport means between the monitoring, controlling or managing units and monitored, controlled or operated power network element or electrical equipment using a wired telecommunication network or a data transmission bus
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/12—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load
    • H02J3/14—Circuit arrangements for ac mains or ac distribution networks for adjusting voltage in ac networks by changing a characteristic of the network load by switching loads on to, or off from, network, e.g. progressively balanced loading
    • H02J3/144—Demand-response operation of the power transmission or distribution network

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Power Engineering (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Public Health (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Water Supply & Treatment (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application relates to a non-invasive load monitoring method, a non-invasive load monitoring apparatus and a computer device. The method comprises the following steps: collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene; inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state; and acquiring the working state of each electric equipment at each sampling time point according to the prediction result. By adopting the method, the monitoring precision can be improved and the calculation cost can be reduced.

Description

Non-invasive load monitoring method, device and computer equipment
Technical Field
The present disclosure relates to the field of power grid monitoring technologies, and in particular, to a non-invasive load monitoring method, apparatus, and computer device.
Background
The grid operator may identify the user's behavior patterns by collecting electricity data (e.g., current, voltage, power, and power usage, etc.) from individual users. With these modes, appropriate strategies can be implemented to improve power efficiency. The key technology for identifying the user behavior mode is non-invasive load monitoring, and under the condition that the privacy of the user is not violated, the electricity utilization data of the user are decomposed into signals of each electric equipment, so that the behavior model of the user is identified.
Conventional non-invasive load monitoring includes probabilistic model-based non-invasive load monitoring and machine learning-based non-invasive load monitoring. However, the conventional monitoring method has problems of high calculation cost and low recognition accuracy.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a non-invasive load monitoring method, apparatus, and computer device that can reduce the computational cost and improve the monitoring accuracy.
In a first aspect, the present application provides a non-invasive load monitoring method comprising:
collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene;
inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state;
and acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
In one embodiment, inputting load data into the predictive model includes: pre-training a double convolution neural network architecture to obtain a prediction model; inputting load data into a prediction model;
the pre-training double convolution neural network architecture, the obtaining of the prediction model comprises:
acquiring training data, wherein the training data comprises historical load data under a preset electric field scene;
extracting an actual power state set and an actual power state sequence according to the historical load data;
inputting training data into a double convolution neural network architecture comprising a first convolution neural network and a second convolution neural network, wherein the first convolution neural network outputs predicted power state distribution, and the second convolution neural network outputs predicted power state corresponding to power consumption; the predicted power state is an element in the actual power state set;
optimizing the double convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence to obtain a predicted model;
the state convolution neural network is a first convolution neural network in the optimized double convolution neural network architecture, and the power convolution neural network is a second convolution neural network in the optimized double convolution neural network architecture.
In one embodiment, optimizing the dual convolutional neural network architecture according to the predicted power state distribution, the predicted power state corresponding power consumption, and the actual power state sequence, the obtaining the prediction model includes:
acquiring a cost function of the double-convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence, and iteratively updating learning parameters in the double-convolution neural network architecture through the cost function until a preset condition is met; obtaining a prediction model;
wherein the cost function includes a power consumption error and an error in the power state distribution.
In one embodiment, the method is characterized by: the power consumption error is determined according to the mean square error between the predicted power consumption and the actual power consumption; the predicted power consumption is determined according to the predicted power state distribution and the power consumption corresponding to the predicted power state, and the actual power consumption is determined according to the actual power state sequence;
the error of the power state distribution is determined according to the predicted power state distribution and the actual power state distribution; the actual power state distribution is determined from the actual power state sequence.
In one embodiment, obtaining the prediction result output by the prediction model includes:
acquiring a first output of a state convolution neural network output;
acquiring a second output of the power convolution neural network output; the first output and the second output are in a matrix form;
and obtaining a prediction result according to the Hadamard product of the first output and the second output.
In one embodiment, according to the prediction result, obtaining the working state of each electric device at each sampling time point includes:
based on the prediction result, acquiring a working state by using a maximized probability distribution method; the working state is represented by the power state of the electric equipment at each sampling time point.
In a second aspect, the present application provides a non-invasive load monitoring apparatus, the apparatus comprising:
the acquisition module is used for acquiring load data corresponding to each sampling time point in the monitoring period under the preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene;
the prediction module is used for inputting the load data into the prediction model and obtaining a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state;
and the output module is used for acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene;
inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state;
and acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
In a fourth aspect, the present application provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:
collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene;
inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state;
and acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
In a fifth aspect, the present application provides a computer program product comprising a computer program which, when executed by a processor, performs the steps of:
collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene;
inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state;
and acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
The non-invasive load monitoring method, the non-invasive load monitoring device and the computer equipment collect load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of each electric equipment at the sampling time point under the preset electric field scene; inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state; and acquiring the working state of each electric equipment at each sampling time point according to the prediction result. The state convolution neural network and the power convolution neural network are matched with each other to mine the coupling relation between the power state and the power state change mode, so that the identification accuracy of the working state of the electric equipment is effectively improved. Meanwhile, when the electric equipment is newly added, the method has higher working state identification precision compared with the traditional technical scheme. In addition, compared with the traditional non-invasive load monitoring method based on the probability model, the method and the device can remarkably reduce the calculation cost and effectively improve the load monitoring efficiency.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related art, the drawings that are required to be used in the embodiments or the related technical descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to the drawings without inventive effort for a person having ordinary skill in the art.
FIG. 1 is a diagram of an application environment for a non-invasive load monitoring method in one embodiment;
FIG. 2 is a flow chart of a non-invasive load monitoring method in one embodiment;
FIG. 3 is a schematic diagram of a structure of a double convolutional neural network architecture in one embodiment;
FIG. 4 is a block diagram of a non-invasive load monitoring apparatus in one embodiment;
fig. 5 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The non-invasive load monitoring method provided by the embodiment of the application can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices, and portable wearable devices, where the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, and the like. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.
In an exemplary embodiment, as shown in fig. 2, a non-invasive load monitoring method is provided, which is illustrated as applied to the terminal 102 in fig. 1, and includes the following steps 202 to 206. Wherein:
step 202, collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprises the total power of all electric equipment in a preset electric field scene at a sampling time point.
The electricity utilization scene can be a household scene or an industrial scene. Correspondingly, each electric equipment can be a household scene or industrial equipment. The range of the electric field scene is divided according to actual needs. For example, taking a home scenario as an example, the electric device may be all electric devices in a single home, or may be all electric devices in all households in a region.
The load data can be acquired through intelligent ammeter acquisition. The load data comprises power, namely, the sum of the power of all electric equipment under the preset electric field scene is collected at each sampling time point in the monitoring period. The load data may be represented asWherein T is E N + And each element in X represents the sum of the power of all electric equipment at the corresponding sampling time point in the T period.
Step 204, inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state.
The prediction model comprises two convolutional neural networks (Convolutional Neural Network, CNN), one is a state convolutional neural network, and the other is a power convolutional neural network, which are respectively used for capturing power state distribution of electric equipment and obtaining corresponding power consumption of the power state.
Suppose that the electric field scene comprisesN electric devices, the sampling interval is 1s, then the power consumption of the ith electric device is expressed asWherein i is E [ N ]]Set [ N ]]Defined as {1,2, …, N }. The total power and the power of the single electric equipment are as follows:
wherein ε t Representing the error.
The power consumption of a common powered device exhibits a clear power state pattern and transitions when the time step after another power state changes drastically, the signal of which is clearly separated from the previous state. For example, for the same air conditioner, the power operated in the normal operation mode is different from the power operated in the power saving mode. Thus, the present embodiment may represent the operating state of the operating device by the power state.
For electric equipment i, set M i ∈N + Indicating the number of power states thereof. In principle, each power state has a unique power level, and at each time step the consumer will be in one power state
In fact, this assumption may be overly simplified, as the actual power readings of the powered device in one power state are subject to disturbances, and thus may have small fluctuations at a fixed power level. In view of the above, this embodiment adoptsRepresenting the power consumption of device i in power state j and at a given time t. Because of the uncertainty in the learning process, it is difficult to pinpoint the individual states in which the device is running at a given time, but rather the probability that device i is in power state j at time tThat is, the power of a single powered device at time t is determined by all M i Power state corresponds to power consumptionAnd probability distribution->To determine, wherein->Is probability ofI.e. the power state of consumer i at a given time t +.>Is the conditional probability of s. Based on the above settings, the power consumption of a single powered device may be considered as a desire for a corresponding power consumption for each power state, formulated as:
the prediction model inputs load data X, and the output prediction result is the power consumption Y of each electric equipment i 。
And 206, acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
And according to the prediction result output by the prediction model, the working state of the electric equipment at each sampling time point can be analyzed and obtained.
The non-invasive load monitoring method disclosed by the embodiment of the invention has the advantages that the state convolutional neural network and the power convolutional neural network are matched with each other, the coupling relation between the power state and the power state change mode is mined, and compared with the non-invasive load monitoring method based on machine learning, the working state recognition precision is effectively improved. Meanwhile, compared with a non-invasive load monitoring method based on a probability model, in a large-range electric field scene containing multiple power state electric equipment, the calculation cost can be greatly reduced, and the identification accuracy is effectively improved when the working state of the newly added electric equipment is identified.
In one embodiment, inputting load data into the predictive model includes: pre-training a double convolution neural network architecture to obtain a prediction model; load data is input into the predictive model. The pre-training double convolution neural network architecture, the obtaining of the prediction model comprises: acquiring training data, wherein the training data comprises historical load data under a preset electric field scene; extracting an actual power state set and an actual power state sequence according to the historical load data; inputting training data into a double convolution neural network architecture comprising a first convolution neural network and a second convolution neural network, wherein the first convolution neural network outputs predicted power state distribution, and the second convolution neural network outputs predicted power state corresponding to power consumption; the predicted power state is an element in the actual power state set; optimizing the double convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence to obtain a predicted model; the state convolution neural network is a first convolution neural network in the optimized double convolution neural network architecture, and the power convolution neural network is a second convolution neural network in the optimized double convolution neural network architecture.
The predictive model is trained based on a double convolutional neural network architecture. As shown in fig. 3, a structural schematic diagram of the double convolutional neural network architecture is shown. In the figure, the state CNN is a state convolutional neural network, that is, an optimized first convolutional neural network. The power CNN is the power convolutional neural network, namely the optimized second convolutional neural network. Two convolutional neural networks in the doubly convolutional neural network architecture share the same input window.
Training the double convolution neural network architecture, training data of an input sharing input window is firstly required to be acquired. The training data in the scheme adopts historical load data under the preset electric field scene, and the historical load data can be data acquired in the field or stored data. The historical load data is the total power of all electric equipment under the preset electric field scene obtained by sampling in a sampling period. Historical loadThe data is expressed as
After the historical load data is acquired, a group of fixed power states of the electric equipment are extracted from the historical load data to form an actual power state set, wherein the number of elements in the set is M i . From the historical load data, a sequence of actual power states may also be extracted, represented asI.e. the actual power state of each sampling period is +.>
Inputting training data into a double convolutional neural network architecture, a first convolutional neural networkOutputting a predicted power state distribution, i.e. & gt, based on the training data>
Wherein,
second convolutional neural networkPredicting power state corresponding power consumption based on training data output, i.e
Wherein,
based on predicted power state distributionPredicting power state versus power consumption->And optimizing the double convolution neural network architecture by the actual power state sequence to obtain a prediction model.
According to the embodiment, the neural network structure is trained through the collected historical load data, and the obtained prediction model can improve the prediction accuracy. Meanwhile, two convolutional neural networks are adopted to collect the predicted power state distribution and the power consumption corresponding to the predicted power state for training data, so that the coupling relation between the power state and the state change mode can be effectively mined, and the recognition precision is improved.
In one embodiment, optimizing the dual convolutional neural network architecture according to the predicted power state distribution, the predicted power state corresponding power consumption, and the actual power state sequence, the obtaining the prediction model includes: acquiring a cost function of the double-convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence, and iteratively updating learning parameters in the double-convolution neural network architecture through the cost function until a preset condition is met; obtaining a prediction model; wherein the cost function includes a power consumption error and an error in the power state distribution.
In this embodiment, a loss function is established to optimize parameters in the model. In model training, a cost function is a function that measures the difference between model predictions and true results. The cost function can calculate a cost value according to the difference between the model prediction result and the real result, and the model has better prediction capability by minimizing the cost value.
The predicted power state distribution and the power consumption corresponding to the predicted power state represent model prediction results, the actual power state sequence represents actual results, a cost function can be obtained through the three data sets, and then iteration updating is carried out on parameters in the double-convolution neural network architecture according to the cost function until preset conditions are met, iteration is stopped, and the double-convolution neural network architecture at the moment is the pre-trained prediction model.
The cost function mainly updates the learning parameters in the double convolution neural network architecture. Let the cost function of device i be L i Updating the learning parameter theta by using a gradient descent mode, wherein the formula is as follows:
wherein θ t-1 Representing the learning parameters, θ, before updating t The updated learning parameter is represented, and α represents the learning rate.
The preset condition for stopping the iteration can be freely selected according to actual needs. For example, a fixed number of iterations may be set as a stop condition, for example, 1000 iterations. When the specified number of iterations is reached, the algorithm stops updating the parameters. For another example, a threshold is set and updating the parameters is stopped when the improvement of the cost function after each iteration is less than the threshold. For another example, a desired model performance index may be set, such as accuracy over training data, mean square error over the test set, etc. When the model performance reaches the set expectations, the algorithm stops updating the parameters.
Since the present embodiment employs a double convolutional neural network architecture, the cost function is composed of two parts, one part is a power consumption error and the other part is an error of power state distribution.
The double convolution neural network architecture is optimized through the cost function, so that the double convolution neural network architecture can be better fitted with training data, and the method has important significance for improving the accuracy and generalization capability of model prediction.
In one embodiment, the method is characterized by: the power consumption error is determined according to the mean square error between the predicted power consumption and the actual power consumption; the predicted power consumption is determined according to the predicted power state distribution and the power consumption corresponding to the predicted power state, and the actual power consumption is determined according to the actual power state sequence; the error of the power state distribution is determined according to the predicted power state distribution and the actual power state distribution; the actual power state distribution is determined from the actual power state sequence.
The power consumption error is determined according to the mean square error between the predicted power consumption and the actual power consumption, the power consumption of the single electric equipment can be regarded as the expectation of the corresponding power consumption of each power state, and the expression for obtaining the power consumption error is as follows:
wherein,predicted power state distribution for the ith device, for>Power consumption for the predicted power state of the ith device,/->Representing the actual power consumption of the i-th device. The actual power consumption is determined according to the actual power state sequence, and the actual power consumption can be obtained by multiplying the actual power state sequence by the residence time of each actual power state and accumulating the multiplied residence time of each actual power state because the power consumption shows an obvious power state mode and the signal transition between each power state is obvious.
In addition, the power consumption error may be determined based on a root mean square error between the predicted power consumption and the actual power consumption.
The error of the power state distribution is determined from the predicted power state distribution and the actual power state distribution, expressed as:
wherein, the reality is thatInter-power state distributionIs determined from the actual sequence of power states.
Finally, the cost function consists of two parts:
in one embodiment, obtaining the prediction result output by the prediction model includes: acquiring a first output of a state convolution neural network output; acquiring a second output of the power convolution neural network output; the first output and the second output are in a matrix form; and obtaining a prediction result according to the Hadamard product of the first output and the second output.
According to the fact that the state convolution neural network is the optimized first convolution neural network, the first output is the predicted power state distributionAccording to the fact that the power convolution neural network is an optimized second convolution neural network, the second output is the predicted power state corresponding power consumption +.>The outputs of the two convolutional neural networks are multiplied element by element, i.e. Hadamard product, to obtain the prediction result +.>The notation is:
in one embodiment, according to the prediction result, obtaining the working state of each electric device at each sampling time point includes: based on the prediction result, acquiring a working state by using a maximized probability distribution method; the working state is represented by the power state of the electric equipment at each sampling time point.
The operating state of the consumer can be obtained by maximizing the probability distribution, expressed as:
the probability distribution of the power states of the electric equipment at different sampling time points is output, and the most probable power states of the electric equipment at the sampling time points can be obtained through the probability distribution, so that the working states of the electric equipment are represented by the power states.
According to the invention, non-invasive load monitoring is performed based on the dual neural network, and the two convolutional neural networks are matched with each other to effectively mine the coupling relation between the power state and the state change mode, so that the recognition accuracy under the scene containing the newly added equipment is improved. Meanwhile, the method disclosed by the invention is an improvement based on a machine learning model, and has the advantage of low calculation cost compared with the traditional method based on a probability model.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiments of the present application also provide a non-invasive load monitoring apparatus for implementing the above-mentioned related non-invasive load monitoring method. The implementation of the solution provided by the device is similar to that described in the above method, so specific limitations in one or more embodiments of the non-invasive load monitoring apparatus provided below may be found in the above limitations of the non-invasive load monitoring method, and will not be described in detail herein.
In one exemplary embodiment, as shown in FIG. 4, a non-invasive load monitoring apparatus is provided, comprising: an acquisition module 402, a prediction module 404, and an output module 406, wherein:
the acquisition module 402 is configured to acquire load data corresponding to each sampling time point in the monitoring period under the preset electric field scene; the load data comprises the total power of all electric equipment in a preset electric field scene at a sampling time point.
The prediction module 404 is configured to input load data into a prediction model, and obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing power state distribution of electric equipment and a power convolution neural network for acquiring corresponding power consumption of the power state.
And the output module 406 is configured to obtain, according to the prediction result, a working state of each electric device at each sampling time point.
In one embodiment, the prediction module 404 is further configured to pre-train the dual convolutional neural network architecture to obtain a prediction model; load data is input into the predictive model.
In one embodiment, the prediction module 404 is further configured to obtain training data, where the training data includes historical load data under a preset electric field scenario; extracting an actual power state set and an actual power state sequence according to the historical load data; inputting training data into a double convolution neural network architecture comprising a first convolution neural network and a second convolution neural network, wherein the first convolution neural network outputs predicted power state distribution, and the second convolution neural network outputs predicted power state corresponding to power consumption; the predicted power state is an element in the actual power state set; optimizing the double convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence to obtain a predicted model; the state convolution neural network is a first convolution neural network in the optimized double convolution neural network architecture, and the power convolution neural network is a second convolution neural network in the optimized double convolution neural network architecture.
In one embodiment, the prediction module 404 is further configured to obtain a cost function of the dual-convolutional neural network architecture according to the predicted power state distribution, the predicted power state corresponding power consumption, and the actual power state sequence, and iteratively update the learning parameters in the dual-convolutional neural network architecture through the cost function until a preset condition is satisfied; obtaining a prediction model; wherein the cost function includes a power consumption error and an error in the power state distribution.
In one embodiment, the power consumption error is determined from a mean square error between the predicted power consumption and the actual power consumption; the predicted power consumption is determined according to the predicted power state distribution and the power consumption corresponding to the predicted power state, and the actual power consumption is determined according to the actual power state sequence; the error of the power state distribution is determined according to the predicted power state distribution and the actual power state distribution; the actual power state distribution is determined from the actual power state sequence.
In one embodiment, the prediction module 404 is further configured to obtain a first output of the state convolutional neural network output; acquiring a second output of the power convolution neural network output; the first output and the second output are in a matrix form; and obtaining a prediction result according to the Hadamard product of the first output and the second output.
In one embodiment, the output module 406 is further configured to obtain the working state by using a maximized probability distribution method based on the prediction result; the working state is represented by the power state of the electric equipment at each sampling time point.
The various modules in the non-invasive load monitoring apparatus described above may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one exemplary embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 5. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured 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, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store load data. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a non-invasive load monitoring method.
In an exemplary embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor performing the steps of all the method embodiments described above when the computer program is executed.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, carries out the steps of all the method embodiments described above.
In an embodiment, a computer program product is provided comprising a computer program which, when executed by a processor, implements the steps of all the method embodiments described above.
It should be noted that, the user information (including, but not limited to, user equipment information, user personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use, and processing of the related data are required to meet the related regulations.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application shall be subject to the appended claims.

Claims (10)

1. A method of non-invasive load monitoring, the method comprising:
collecting load data corresponding to each sampling time point in a monitoring period under a preset electric field scene; the load data comprise the total power of each electric equipment at the sampling time point under the preset electric field scene;
inputting the load data into a prediction model to obtain a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing the power state distribution of the electric equipment and a power convolution neural network for acquiring the corresponding power consumption of the power state;
and acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
2. The method of claim 1, wherein said inputting the load data into a predictive model comprises: pre-training a double convolution neural network architecture to obtain the prediction model; inputting the load data into the predictive model;
the pre-training double convolution neural network architecture, the obtaining the prediction model comprises:
acquiring training data, wherein the training data comprises historical load data under the preset electric field scene;
extracting an actual power state set and an actual power state sequence according to the historical load data;
inputting the training data into the double convolutional neural network architecture comprising a first convolutional neural network and a second convolutional neural network, wherein the first convolutional neural network outputs a predicted power state distribution, and the second convolutional neural network outputs a predicted power state corresponding to power consumption; the predicted power state is an element in the set of actual power states;
optimizing the double convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence to obtain the prediction model;
the state convolution neural network is the first convolution neural network in the optimized double convolution neural network architecture, and the power convolution neural network is the second convolution neural network in the optimized double convolution neural network architecture.
3. The method of claim 2, wherein optimizing the dual convolutional neural network architecture according to the predicted power state distribution, the predicted power state corresponding power consumption, and the actual power state sequence, the obtaining the prediction model comprises:
acquiring a cost function of the double-convolution neural network architecture according to the predicted power state distribution, the power consumption corresponding to the predicted power state and the actual power state sequence, and iteratively updating learning parameters in the double-convolution neural network architecture through the cost function until a preset condition is met; acquiring the prediction model;
wherein the cost function includes a power consumption error and an error in a power state distribution.
4. A method according to claim 3, characterized in that: the power consumption error is determined according to the mean square error between the predicted power consumption and the actual power consumption; the predicted power consumption is determined according to the predicted power state distribution and the power consumption corresponding to the predicted power state, and the actual power consumption is determined according to the actual power state sequence;
the error of the power state distribution is determined according to the predicted power state distribution and the actual power state distribution; the actual power state distribution is determined from the actual power state sequence.
5. The method of claim 1, wherein the obtaining the prediction result output by the prediction model comprises:
acquiring a first output of the state convolution neural network output;
acquiring a second output of the power convolution neural network output; the first output and the second output are both in a matrix form;
and obtaining the prediction result according to the Hadamard product of the first output and the second output.
6. The method of claim 1, wherein the obtaining, according to the prediction result, an operating state of each of the electric devices at each of the sampling time points includes:
based on the prediction result, acquiring the working state by using a maximized probability distribution method; the working state is represented by the power state of the electric equipment at each sampling time point.
7. A non-invasive load monitoring apparatus, the apparatus comprising:
the acquisition module is used for acquiring load data corresponding to each sampling time point in the monitoring period under the preset electric field scene; the load data comprise the total power of each electric equipment at the sampling time point under the preset electric field scene;
the prediction module is used for inputting the load data into a prediction model and obtaining a prediction result output by the prediction model; the prediction model comprises a state convolution neural network for capturing the power state distribution of the electric equipment and a power convolution neural network for acquiring the corresponding power consumption of the power state;
and the output module is used for acquiring the working state of each electric equipment at each sampling time point according to the prediction result.
8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
CN202311527771.1A 2023-11-15 2023-11-15 Non-invasive load monitoring method, device and computer equipment Pending CN117578487A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311527771.1A CN117578487A (en) 2023-11-15 2023-11-15 Non-invasive load monitoring method, device and computer equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311527771.1A CN117578487A (en) 2023-11-15 2023-11-15 Non-invasive load monitoring method, device and computer equipment

Publications (1)

Publication Number Publication Date
CN117578487A true CN117578487A (en) 2024-02-20

Family

ID=89892899

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311527771.1A Pending CN117578487A (en) 2023-11-15 2023-11-15 Non-invasive load monitoring method, device and computer equipment

Country Status (1)

Country Link
CN (1) CN117578487A (en)

Similar Documents

Publication Publication Date Title
Wen et al. Big data driven marine environment information forecasting: a time series prediction network
Ma et al. A hybrid attention-based deep learning approach for wind power prediction
CN112766596B (en) Construction method of building energy consumption prediction model, energy consumption prediction method and device
Razghandi et al. Short-term load forecasting for smart home appliances with sequence to sequence learning
CN111027672A (en) Time sequence prediction method based on interactive multi-scale recurrent neural network
Dong et al. Forecasting smart meter energy usage using distributed systems and machine learning
CN115034430A (en) Carbon emission prediction method, device, terminal and storage medium
Wang et al. A Transformer-based multi-entity load forecasting method for integrated energy systems
CN117472511A (en) Container resource monitoring method, device, computer equipment and storage medium
CN115577647B (en) Power grid fault type identification method and intelligent agent construction method
CN115952928B (en) Short-term power load prediction method, device, equipment and storage medium
CN117056776A (en) Carbon emission monitoring method, apparatus, device, medium and computer program product
CN116523001A (en) Method, device and computer equipment for constructing weak line identification model of power grid
CN115130620B (en) Power equipment power utilization mode identification model generation method and device
CN117034046A (en) Flexible load adjustable potential evaluation method based on ISODATA clustering
CN116011655A (en) Load ultra-short-term prediction method and system based on two-stage intelligent feature engineering
CN117578487A (en) Non-invasive load monitoring method, device and computer equipment
Wei et al. Short-term load forecasting using spatial-temporal embedding graph neural network
CN115081551A (en) RVM line loss model building method and system based on K-Means clustering and optimization
CN115481767A (en) Operation data processing method and device for power distribution network maintenance and computer equipment
CN113741402A (en) Equipment control method and device, computer equipment and storage medium
Li et al. Comparison and application potential analysis of autoencoder-based electricity pattern mining algorithms for large-scale demand response
CN116526582B (en) Combined dispatching method and system for electric power unit based on artificial intelligence combined driving
CN117313952A (en) Load prediction method, device, equipment and storage medium
Meng et al. An Ensemble Learning-based Short-Term Load Forecasting on Small Datasets

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination