CN114881310A - Energy consumption prediction device and method - Google Patents
Energy consumption prediction device and method Download PDFInfo
- Publication number
- CN114881310A CN114881310A CN202210448324.6A CN202210448324A CN114881310A CN 114881310 A CN114881310 A CN 114881310A CN 202210448324 A CN202210448324 A CN 202210448324A CN 114881310 A CN114881310 A CN 114881310A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- real
- power curve
- historical
- time
- 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
Links
- 238000005265 energy consumption Methods 0.000 title claims abstract description 93
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 47
- 238000012360 testing method Methods 0.000 claims abstract description 37
- 230000005611 electricity Effects 0.000 claims abstract description 16
- 239000013598 vector Substances 0.000 claims description 16
- 238000001514 detection method Methods 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims 2
- 210000002569 neuron Anatomy 0.000 description 15
- 230000006870 function Effects 0.000 description 14
- 238000005259 measurement Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Economics (AREA)
- Mathematical Analysis (AREA)
- Software Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Algebra (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Databases & Information Systems (AREA)
- Quality & Reliability (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Biomedical Technology (AREA)
Abstract
The invention discloses an energy consumption prediction device and method. Wherein the method comprises: distinguishing the period types of the historical data set according to a plurality of electricity utilization periods; configuring a base model corresponding to the time period type; a first training set according to the period type; trained respective said base models; training a meta-model according to the test energy consumption labels of the respective base models and the real second energy consumption label; combining each base model and meta model into a combined model; acquiring a real-time data set of at least one real-time power curve; and acquiring energy consumption prediction of the real-time data set according to the combined model.
Description
Technical Field
The invention relates to the technical field of electric power, in particular to an energy consumption prediction device and method applied to energy consumption evaluation.
Background
In the prior art, the statistics of household energy consumption still depends on meter reading data of a meter. However, in long-time meter reading statistics, due to external or self factors, the meter is unnecessarily prevented from generating errors. The existing metering error is large, the accuracy is improved by using a deep learning algorithm, but the accuracy is high, the historical data is excessively depended on, the current historical data does not consider a complex household electricity environment, and meanwhile, the adjustment cannot be made according to the change of the household electricity.
Disclosure of Invention
Based on the above, the embodiment of the invention aims to solve the problem that in the prior art, the prediction methods such as deep learning and the like depend on historical samples and prediction incapacity is caused by not considering the complex environment of household electricity utilization.
At least one energy consumption prediction device and method are disclosed. Based on the disclosure of the present embodiment,
the invention discloses an energy consumption prediction method on one hand.
The method comprises the following steps:
obtaining a first historical data set of at least one first historical power curve and at least one associated first energy consumption label;
differentiating a period type of the first historical data set according to a plurality of electricity usage periods;
obtaining a first combination of at least a first eigenvector comprising the first historical data set and the first energy consumption label;
configuring a base model corresponding to the time period type;
creating a first training set from a number of the first combinations of the period types;
training the respective base models according to the first training set to obtain the respective trained base models;
obtaining a second historical data set of at least one second historical power curve and at least one associated second energy consumption label;
obtaining a second combination of at least a second eigenvector comprising the second historical dataset and the second energy consumption label;
creating a test set according to a plurality of second combinations;
testing each base model according to the second combination of the test set to obtain a test energy consumption label of each base model of each second combination;
acquiring a third combination comprising a test energy consumption label acquired by the second combination in each base model respectively and the second energy consumption label;
creating a second training set according to a plurality of third combinations;
training a configured meta-model according to the second training set;
combining the base model and the meta model into a combined model;
acquiring a real-time data set of at least one real-time power curve;
and acquiring energy consumption prediction of the real-time data set according to the combined model.
The invention discloses an energy consumption prediction device on the other hand.
The device comprises:
the first data acquisition module is used for acquiring a first historical data set of at least one first historical power curve and at least one associated first energy consumption label; differentiating a period type of the first historical data set according to a plurality of electricity usage periods; the first data acquisition module is used for acquiring at least one first feature vector comprising the first historical data set and a first combination of the first energy consumption label; creating a first training set from a number of the first combinations of the period types;
the base model module is used for acquiring a base model corresponding to the time interval type; training the respective base models according to the first training set to obtain the respective trained base models;
the test acquisition module is used for acquiring a second historical data set of at least one second historical power curve and at least one associated second energy consumption label; obtaining a second combination of at least a second eigenvector comprising the second historical dataset and the second energy consumption label; creating a test set according to a plurality of second combinations;
the second training module is used for testing each base model according to the second combination of the test set to obtain a test energy consumption label of each base model of each second combination; acquiring a third combination comprising a test energy consumption label acquired by the second combination in each base model respectively and the second energy consumption label; creating a second training set according to a plurality of third combinations;
the integrated model module is used for training a configured meta-model according to the second training set; combining the base model and the meta model into a combined model;
the prediction module is used for acquiring a real-time data set of at least one real-time power curve; and acquiring energy consumption prediction of the real-time data set according to the combined model.
In view of the above, other features and advantages of the disclosed exemplary embodiments will become apparent from the following detailed description of the disclosed exemplary embodiments, which proceeds with reference to the accompanying drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flow chart of a method of an embodiment for obtaining a first training set;
FIG. 2 is a flow chart of a method training a base model and obtaining a second training set in an embodiment;
FIG. 3 is a flow diagram of a method training meta-model and combined joint model in an embodiment;
FIG. 4 is a flow chart of a method for obtaining a first feature vector in an embodiment;
FIG. 5 is a flow chart of a method for optimizing weights of a base model in an embodiment;
FIG. 6 is a flowchart illustrating the method for determining similarity between curves according to the embodiment.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
Fig. 1 is a schematic flow chart of an energy consumption prediction method provided in an embodiment, and as shown in fig. 1 to fig. 3, the method provided in this embodiment is suitable for predicting energy consumption of household electricity. The method can be realized in a software and/or hardware mode and is configured in the server. The method specifically comprises the following steps:
s101, a plurality of first historical power curves representing household electricity are obtained. The first historical power curve is used to characterize the instantaneous power change captured by hardware measurement in a unit hour, and this instantaneous power capture can be based on the measurement of the instantaneous voltage and current.
S102, a first historical data set of a first historical power curve and a first energy consumption label actually obtained are obtained. The first energy consumption tag is the acquisition of real electricity usage for the first historical power curve. The difference between the real electricity consumption and the metering electricity consumption obtained by integrating the instantaneous power and time is large, such as self error of a metering device, and error of the metering device influenced by environmental factors such as voltage, current, temperature, frequency and the like.
Furthermore, the energy consumption tag may be configured for quantitative energy consumption values and may also be configured for qualitative energy consumption levels. The predicted result of the method in this embodiment is differentiated according to the definition of the energy consumption label. Optionally, the energy consumption label in this embodiment is characterized as an energy consumption value.
And S103, distinguishing the time period types of the first historical data sets according to different electricity utilization time periods. The power consumption time intervals are mainly distinguished based on the power consumption characteristics of the household in different time intervals, such as the power consumption time interval for the household to use power electric appliances in a centralized manner from 6 to 9 in the morning, the power consumption time interval for the household to prepare meals for the household from 11 to 13 in the noon, 17 to 19 in the evening, the power consumption time interval for using kitchen electric appliances, the current time interval for the household to use high-power domestic electric appliances in a centralized manner from 20 to 22 in the night, and the like.
Wherein the time segment type determination for the first historical data set may be obtained by a prior tag or comparison to a similarity of a standard curve corresponding to the time segment type.
For example, a part of the first historical power curve is represented by a column vector Xk, k represents the length of the time series of the curve, n column vectors are combined into a matrix X ═ Xk1, Xk2,. Xkn ], so that a similarity coefficient S represents the similarity degree of two curve series, generally, the larger the value of S is, the greater the correlation degree of the two curve series is, the similarity coefficient S can be configured to be,
wherein, Xa and Xb represent two curve time series, Sab represents the correlation coefficient between Xa and Xb, Cov (Xa and Xb) represents the covariance between Xa and Xb, and D (Xa) and D (Xb) represent the variance between Xa and Xb.
S104, a plurality of first feature vectors comprising the first historical data set and a first combination of the first energy consumption labels are obtained.
Fig. 4 shows steps S201 to S205 for acquiring the first feature vector.
S201, acquiring frequency domain waveform data of the first historical data set according to Fourier transform.
Wherein the first historical power curve is an instantaneous power change over a period of time, and belongs to an energy-limited signal. The first historical power curve may utilize a fourier transform to acquire frequency domain waveform data. The frequency domain waveform data characterizes the frequency components of the instantaneous power change, as well as the amplitude of each frequency component, i.e., the power spectrum.
S202, all frequency components and the maximum amplitude of each frequency component are obtained from the frequency domain waveform data.
S203, a plurality of frequency components with the maximum amplitude intensity arranged in the front are selected, and normalization processing is carried out on the frequency components.
S204, a variance of the set of the plurality of normalized frequency components is obtained as a first eigenvector x 1.
The variance characterization of the frequency component set is the average of the squared differences between each element value in the set and the average of the overall element values, i.e. the deviation degree of each element value in the set. The plurality of frequency components is used to measure the degree of deviation of the dominant frequency components of the first historical power curve.
S205, an integrated area of the first historical power curve is obtained as another first feature vector x 2.
And S105, configuring a plurality of base models corresponding to the time interval types.
The base model may be a bp (back propagation) nerve base model. Specifically, the BP neural network is a multilayer feedforward network trained according to error back propagation, and the BP neural base model can realize the function of predicting the power consumption according to the first feature vector.
For example, as can be seen in FIG. 5, the base model consists of an input layer, a hidden layer, and an output layer; and the S-shaped transfer function is chosen,by passing back the error function, the error signal,wherein Ti is the expected output, Oi is the calculated output of the network, and the error function E is minimized by adjusting the weight and the threshold of the network.
Specifically, then, the input layer consists of 2 neurons, i.e., 2 variable nodes. The variable nodes are x1 and x 2. The output layer consists of 1 neuron, i.e. 1 variable node. In addition, the hidden layer has one and only one layer, and the number of neurons in the hidden layer refers to an empirical formulaWherein n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is the interval [1, 10 ]]Is constant over time. Thus, the value of a in the example in the embodiment is 4; then there are 5 neurons in the hidden layer. Meanwhile, in the embodiment, the S-type tangent function tansig is selected as the excitation function of the hidden layer neuron, and the energy consumption can be predicted through the basic model.
In this embodiment, the basic model configurations corresponding to different time period types are the same, and the difference is the input of different time period types for training each basic model.
S106, a first training set is created according to a plurality of first combinations of different time period types. The first combined number for each different time period type in the first training set remains the same.
Optionally, the similarity of the first historical curve corresponding to each first combination in different time period types is determined based on the similarity coefficient S, so as to ensure the difference of the first combinations in different time period types.
S107, training the respective base models according to the first training set to obtain the trained respective base models. In particular, the base models for different epoch types are trained with a first combination of respective epoch types. The weights of the trained base models at different time periods are not necessarily identical, and a basis is provided for the ensemble learning in the embodiment.
S108, a second historical data set of at least one second historical power curve and at least one associated second energy consumption label are obtained.
S109, a second combination of at least one second feature vector and a second energy consumption label including a second historical data set is obtained. The second eigenvector is kept the same as the first eigenvector, and the second energy consumption label is also characterized by the real energy consumption corresponding to the second historical power curve.
And S110, creating a test set based on a plurality of second combinations without distinguishing the period types.
And S111, respectively and equally testing each trained base model according to the second combination of the test set to obtain the test energy consumption label of each second combination on each base model, namely predicting the obtained test energy consumption. Then the second combination of each test set would be entered as multiple inputs into the base model and multiple times would result in test energy consumption signatures that are not necessarily the same.
And S112, acquiring a plurality of third combinations including the test energy consumption labels and the second energy consumption labels respectively acquired by the second combination in each base model.
And S113, creating a second training set according to the obtained plurality of third combinations.
And S114, training a configured meta model according to the second training set. Where the meta-model is a stack fusion in application ensemble learning.
Alternatively, the meta-model may be a bp (backpropagation) nerve-based model or the like. Specifically, the BP neural network is a multi-layer feedforward network trained according to error back propagation, and the BP neural base model can realize the function of predicting the power consumption based on the prediction results of a plurality of base models.
For example, a meta-model is also composed of an input layer, a hidden layer, and an output layer; and the S-shaped transfer function is chosen,by passing back the error function, the error signal,wherein Ti is the expected output, Oi is the calculated output of the network, and the error function E is minimized by adjusting the weight and the threshold of the network.
In particular, then, the input layerThe model prediction method is characterized by comprising neurons with the number equal to that of the base model, wherein each neuron corresponds to a variable node, namely a prediction result of the base model. The output layer consists of 1 neuron, i.e. 1 variable node. In addition, the hidden layer has one and only one layer, and the number of neurons in the hidden layer refers to an empirical formulaWherein n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is the interval [1, 10 ]]Is constant over time. Thus, the value of a in the examples is a random value of 4 to 6. Meanwhile, in the embodiment, the S-type tangent function tansig is selected as the excitation function of the hidden layer neuron, and the energy consumption can be predicted through the neuron model.
And S115, combining and packaging the plurality of base models and meta models into an integrated learning combined model.
The combined model is provided with the same number of input ends as the basic model, each input end can input the same power data set, and the combined model predicts the corresponding energy consumption according to the input power data sets.
The real-time energy consumption prediction step of the present embodiment is shown in fig. 4 based on the configured combined model.
S201, a real-time power curve is obtained.
S202, acquiring a real-time data set of the real-time power curve.
S203, selecting the packaged combined model to predict the energy consumption of the real-time data set.
Fig. 5 shows that, in this embodiment, when the household electricity consumption habits change, such as an increase in the number of people, a change in work and rest, and the like, and the real-time power curve and the historical power curve drift, and the joint model prediction fails, the base model and the meta model may be iterated in an incremental learning manner.
S301, judging that data drifting exists between the acquired real-time power curve and any historical power curve in the same time period type.
Wherein, the data drift is judged by measuring the curve similarity of the real-time power curve and any historical power curve in the same time period type based on the relative entropy detection, namely
P represents a function of the first historical data set and Q represents the real-time data set. Then the curve similarity is greater than or equal to a distance threshold and data drift is determined to exist.
S302, after the fact that the real-time power curve has data drift is judged, a first feature vector of the real-time data set is obtained and enters a first combination of the corresponding time interval type together with the iteration of energy consumption prediction replacement predicted in real time, and therefore the weight of the base model of the corresponding time interval type is re-optimized.
It is necessary to copy the first feature vector of the real-time data set a plurality of times and to iterate a first combination of a plurality of corresponding session types, respectively, in combination with the energy consumption prediction.
Preferably, the measurement of the data drift can be based on the drift detection step of the data window shown in fig. 6.
S401, creating a data window with at least one time span of [ T0, T1], initializing the starting time T0 of the data window, and defining the span step Tx of the data window along the time axis.
S402, acquiring a plurality of first window integral areas of the data window in the real-time power curve step, and combining the plurality of first window integral areas to form a first area data set.
S403, acquiring a plurality of second window integral areas of the data window in the step of the historical power curve, and combining the plurality of second window integral areas to form a second area data set.
S404, similarity is measured according to difference of the variances of the first area data set and the second area data set.
In this respect, the respective area data set characterizes the power variation of the respective power curve in the time series. The variance of each area data can be used for measuring the deviation degree of power change, and further the similarity between curves can be measured.
Preferably, the larger the time span from the time T0 to the time T1 is, the smaller the span step Tx is, and the measurement accuracy of the curve similarity can be effectively improved.
The embodiment of the invention at least discloses an energy consumption prediction device. The device comprises a first data acquisition module, a base model module and a first data acquisition module, wherein the first data acquisition module is used for acquiring a first historical data set of at least one first historical power curve and at least one associated first energy consumption label; distinguishing a period type of the first historical data set according to a plurality of electricity utilization periods; the first data acquisition module is used for acquiring at least one first feature vector comprising a first historical data set and a first combination of a first energy consumption label; a first training set is created from a number of first combinations of period types. The base model module is used for acquiring a base model corresponding to the configuration and the time interval type; and training the respective base models according to the first training set to obtain the trained respective base models. The test acquisition module is used for acquiring a second historical data set of at least one second historical power curve and at least one associated second energy consumption label; obtaining a second combination of at least one second eigenvector comprising a second historical data set and a second energy consumption label; a test set is created from the number of second combinations. The second training module is used for testing each base model according to the second combination of the test set to obtain a test energy consumption label of each second combination on each base model; acquiring a third combination of a test energy consumption label and a second energy consumption label which are respectively acquired by the second combination in each base model; a second training set is created from a number of third combinations. The integrated model module is used for training a configured meta-model according to the second training set; the combined base model and the meta model are combined models. The prediction module is used for acquiring a real-time data set of at least one real-time power curve; and acquiring energy consumption prediction of the real-time data set according to the combined model.
Of course, the storage medium provided by the embodiments of the present invention contains computer-executable instructions, and the computer-executable instructions are not limited to the above method operations, and may also perform related operations of any embodiments of the present invention.
From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly, can also be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as a floppy disk, a Read-only memory (ROM), a Random Access Memory (RAM), a FLASH memory (FLASH), a hard disk or an optical disk of a computer, and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device) to execute the methods according to the embodiments of the present invention.
It should be noted that, in the embodiment of the above search apparatus, each included unit and module are merely divided according to functional logic, but are not limited to the above division as long as the corresponding functions can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
Claims (10)
1. A method of energy consumption prediction, the method comprising:
obtaining a first historical data set of at least one first historical power curve and at least one associated first energy consumption label;
differentiating a period type of the first historical data set according to a plurality of electricity usage periods;
obtaining a first combination of at least a first eigenvector comprising the first historical data set and the first energy consumption label;
configuring a base model corresponding to the time period type;
creating a first training set from a number of the first combinations of the period types;
training the respective base models according to the first training set to obtain the respective trained base models;
obtaining a second historical data set of at least one second historical power curve and at least one associated second energy consumption label;
obtaining a second combination of at least a second eigenvector comprising the second historical dataset and the second energy consumption label;
creating a test set according to a plurality of second combinations;
testing each base model according to the second combination of the test set to obtain a test energy consumption label of each base model of each second combination;
acquiring a third combination comprising a test energy consumption label acquired by the second combination in each base model respectively and the second energy consumption label;
creating a second training set according to a plurality of third combinations;
training a configured meta-model according to the second training set;
combining the base model and the meta model into a combined model;
acquiring a real-time data set of at least one real-time power curve;
and acquiring energy consumption prediction of the real-time data set according to the combined model.
2. The method of energy consumption prediction according to claim 1,
and cleaning at least a pseudo-catastrophe point of the historical power curve when the historical power curve is obtained, wherein the pseudo-catastrophe point is not related to energy consumption prediction at least.
3. The energy consumption prediction method of claim 2,
cleaning the pseudo-mutation points according to a local outlier factor detection method;
Wherein m is the number of the object to be measured and the points from near to far, o' is the neighborhood point of the object to be measured, Nm (o) is the m-th distance neighborhood of the object o, and Ird (o) is the local reachable density.
4. The method of energy consumption prediction according to claim 1,
obtaining the first feature vector, configured to:
obtaining frequency domain waveform data of the first historical data set according to at least a fourier transform;
acquiring different frequency components according to the frequency domain waveform data;
selecting a part of the frequency components with amplitude intensity arranged in front, and normalizing the obtained frequency components;
and acquiring the variance of a plurality of normalized frequency components as at least one first feature vector.
5. The energy consumption prediction method of claim 4,
and acquiring the integral area of the first historical power curve as at least one first characteristic vector.
6. The energy consumption prediction method of claim 4,
the method comprises the following steps:
judging that data drift exists between the real-time power curve and any historical power curve in the same time period type;
and after detecting that the real-time power curve has data drift, enabling the real-time data set and the energy consumption prediction to iterate at least one first combination corresponding to the time interval type.
7. The energy consumption prediction method of claim 6,
determining the data drift, configured to:
measuring the curve similarity of the real-time power curve and any historical power curve in the same time period type based on relative entropy detection;
and judging that data drift exists when the curve similarity of the real-time power curve and at least one historical power curve is greater than or equal to a distance threshold.
8. The energy consumption prediction method of claim 6,
the measure of the similarity of the curves is configured to,
creating a data window with at least one time span of [ T0, T1], initializing a starting time T0 of the data window, and defining a span step Tx of the data window along a time axis;
acquiring a plurality of first window integral areas of the data window in the real-time power curve step, and combining the plurality of first window integral areas to form a first area data set;
acquiring a plurality of second window integral areas of the data window in the step of the historical power curve, and combining the plurality of second window integral areas to form a first area data set;
measuring the similarity according to the difference of the variance of the first area data set and the second area data set.
9. The energy consumption prediction method of claim 6,
when data drift exists between the real-time power curve and most of the historical power curves in the same time period type, detecting whether concept drift exists in the real-time power curve;
and after detecting that the real-time power curve has concept drift, enabling the real-time data set and the energy consumption prediction iteration to completely correspond to the first combination of the time interval types.
10. An energy consumption prediction apparatus, characterized in that,
the device comprises:
the first data acquisition module is used for acquiring a first historical data set of at least one first historical power curve and at least one associated first energy consumption label; differentiating a period type of the first historical data set according to a plurality of electricity usage periods; the first data acquisition module is used for acquiring at least one first feature vector comprising the first historical data set and a first combination of the first energy consumption label; creating a first training set from a number of the first combinations of the period types;
the base model module is used for acquiring a base model corresponding to the time interval type; training the respective base models according to the first training set to obtain the respective trained base models;
the test acquisition module is used for acquiring a second historical data set of at least one second historical power curve and at least one associated second energy consumption label; obtaining a second combination of at least a second eigenvector comprising the second historical dataset and the second energy consumption label; creating a test set according to a plurality of second combinations;
the second training module is used for testing each base model according to the second combination of the test set to obtain a test energy consumption label of each base model of each second combination; acquiring a third combination comprising a test energy consumption label acquired by the second combination in each base model respectively and the second energy consumption label; creating a second training set according to a plurality of third combinations;
the integrated model module is used for training a configured meta-model according to the second training set; combining the base model and the meta model into a combined model;
the prediction module is used for acquiring a real-time data set of at least one real-time power curve; and acquiring energy consumption prediction of the real-time data set according to the combined model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210448324.6A CN114881310A (en) | 2022-04-27 | 2022-04-27 | Energy consumption prediction device and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210448324.6A CN114881310A (en) | 2022-04-27 | 2022-04-27 | Energy consumption prediction device and method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114881310A true CN114881310A (en) | 2022-08-09 |
Family
ID=82672652
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210448324.6A Pending CN114881310A (en) | 2022-04-27 | 2022-04-27 | Energy consumption prediction device and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114881310A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117590763A (en) * | 2024-01-18 | 2024-02-23 | 中网华信科技股份有限公司 | Intelligent park energy data management and control system |
-
2022
- 2022-04-27 CN CN202210448324.6A patent/CN114881310A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117590763A (en) * | 2024-01-18 | 2024-02-23 | 中网华信科技股份有限公司 | Intelligent park energy data management and control system |
CN117590763B (en) * | 2024-01-18 | 2024-03-19 | 中网华信科技股份有限公司 | Intelligent park energy data management and control system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Mengistu et al. | A cloud-based on-line disaggregation algorithm for home appliance loads | |
Aprillia et al. | Statistical load forecasting using optimal quantile regression random forest and risk assessment index | |
Si et al. | Electric load clustering in smart grid: Methodologies, applications, and future trends | |
Faustine et al. | A survey on non-intrusive load monitoring methodies and techniques for energy disaggregation problem | |
Ramos et al. | A data-mining-based methodology to support MV electricity customers’ characterization | |
He et al. | A generic optimisation-based approach for improving non-intrusive load monitoring | |
Zhang et al. | Anomaly detection in premise energy consumption data | |
Ma et al. | Toward energy-awareness smart building: Discover the fingerprint of your electrical appliances | |
US20120290230A1 (en) | Methods and Apparatuses for Monitoring Energy Consumption and Related Operations | |
Chang et al. | Electricity demand analysis using cointegration and error-correction models with time varying parameters: The Mexican case | |
Zeinal-Kheiri et al. | Residential load disaggregation considering state transitions | |
KR102060301B1 (en) | Apparatus and method for analysing living pattern based on power data | |
CN112101110A (en) | Non-invasive load identification method for user side of power system | |
Das et al. | Measuring the degree of non‐stationarity of a time series | |
CN114881310A (en) | Energy consumption prediction device and method | |
Humala et al. | Universalnilm: A semi-supervised energy disaggregation framework using general appliance models | |
Groß et al. | Comparison of stochastic load profile modeling approaches for low voltage residential consumers | |
Rehman et al. | Comparative evaluation of machine learning models and input feature space for non-intrusive load monitoring | |
Rahman et al. | Deep learning based ensemble method for household energy demand forecasting of smart home | |
Zhao et al. | Electricity usage profile disaggregation of hourly smart meter data | |
Solatidehkordi et al. | An IoT deep learning-based home appliances management and classification system | |
Huang et al. | Designing customized energy services based on disaggregation of heating usage | |
Park et al. | Demand power forecasting with data mining method in smart grid | |
Zhang et al. | Forecasting Residential Energy Consumption Using Support Vector Regressions | |
Kommey et al. | An artificial intelligence‐based non‐intrusive load monitoring of energy consumption in an electrical energy system using a modified K‐Nearest Neighbour algorithm |
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 |