CN114386314A - Non-invasive modeling method of comprehensive energy system based on zero sample learning - Google Patents
Non-invasive modeling method of comprehensive energy system based on zero sample learning Download PDFInfo
- Publication number
- CN114386314A CN114386314A CN202111495935.8A CN202111495935A CN114386314A CN 114386314 A CN114386314 A CN 114386314A CN 202111495935 A CN202111495935 A CN 202111495935A CN 114386314 A CN114386314 A CN 114386314A
- Authority
- CN
- China
- Prior art keywords
- equipment
- state
- energy system
- type
- 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
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000004146 energy storage Methods 0.000 claims abstract description 18
- 238000010248 power generation Methods 0.000 claims abstract description 14
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 6
- 230000015654 memory Effects 0.000 claims abstract description 6
- 238000013145 classification model Methods 0.000 claims abstract description 3
- 125000004432 carbon atom Chemical group C* 0.000 claims description 21
- 230000007704 transition Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 8
- 238000000354 decomposition reaction Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 238000005406 washing Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 230000005540 biological transmission Effects 0.000 claims 1
- 238000005094 computer simulation Methods 0.000 abstract description 6
- 238000000605 extraction Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 13
- 238000005516 engineering process Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
-
- 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
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention belongs to the technical field of energy system modeling, and particularly relates to a non-invasive modeling method of a comprehensive energy system based on zero sample learning. The method classifies the equipment in the self-energy source by using zero sample learning based on sparse coding, trains a classification model for a known data set and auxiliary information of the known data set of the comprehensive energy system, and effectively migrates the knowledge learned by the known data set into an unknown data set to realize classification of the unknown equipment data set; extracting the running states of the equipment by adopting a bidirectional long-short term memory network, and dividing the running states of the electric equipment, the power generation equipment and the energy storage equipment in the comprehensive energy system according to the extraction result; and establishing a multi-mode factor hidden Markov model based on the running state of the equipment to realize non-invasive modeling of the comprehensive energy system containing the equipment of unknown type. The method accurately identifies the unknown type equipment in the comprehensive energy system, effectively reduces the average error of the non-invasive modeling method, and obviously improves the precision and the accuracy of the model.
Description
Technical Field
The invention belongs to the technical field of energy system modeling, and particularly relates to a non-invasive modeling method of a comprehensive energy system based on zero sample learning.
Background
The comprehensive energy system is a novel open energy ecosystem which takes clean renewable energy sources such as wind and light and the like as main primary energy sources and takes an energy technology, an advanced control technology, an intelligent optimization technology, an information processing technology and the like as implementation means, can realize the high-efficiency utilization of the renewable energy sources, and improves the occupation ratio of the renewable energy sources in the production and consumption of the primary energy sources. The comprehensive energy system modeling is the basis for researching the safety and stability control and the energy optimization scheduling of the energy Internet system.
At present, domestic and foreign scholars mainly concentrate on the research of invasive modeling methods in the aspect of comprehensive energy system modeling, and the methods respectively model equipment contained in an energy system and integrate all models according to energy types so as to represent the integral model of the energy system. However, this modeling approach requires not only the equipment that the exhaustive system contains, but also the mechanism of operation of each equipment, resulting in a number of difficulties in modeling the energy system, including:
(1) the model is more complex along with the increase of internal equipment of the energy system, and the model parameters and state variables are increased when the model is subjected to optimization control;
(2) due to the nonlinearity and uncertainty of the energy system, a unified model capable of reflecting the state of the energy system is difficult to integrate;
(3) the system operation state is diversified, and different models need to be established and controlled.
The non-invasive load monitoring method provides a new solution for energy system modeling, and refined internal load category and use state data of the user can be obtained by decomposing and identifying total data of the user in the non-invasive load monitoring. However, the types of devices included in the integrated energy system are various, and with the updating of various devices and the access of new devices, the existing method needs to continuously update the system structure and the device model, which greatly affects the precision of the integrated energy system model.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a non-invasive modeling method of a comprehensive energy system based on zero sample learning. The aim is to achieve the aim of the invention which is high in precision and accuracy and can carry out non-invasive modeling on an integrated energy system containing unknown equipment.
The technical scheme adopted by the invention for realizing the purpose is as follows:
a non-intrusive modeling method of an integrated energy system based on zero sample learning comprises the following steps:
step 3, dividing according to the extracted running state;
and 4, step 4: establishing a multi-mode factor hidden Markov model according to the division result;
and 5: performing parameter estimation on the established multi-mode factor hidden Markov model;
step 6: and decoding the hidden state of the multi-mode factor hidden Markov model according to the parameter estimation result, and finally establishing a comprehensive energy system decomposition model.
Further, the known types of devices in the integrated energy system include: photovoltaic power generation equipment, wind power generation equipment, electric energy storage equipment, electric vehicles, rotating electrical machines, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.
Furthermore, the step of classifying the known type equipment and the unknown type equipment is to classify the equipment in the self-energy source by utilizing zero sample learning based on sparse coding, train a classification model by utilizing a known data set of the comprehensive energy system and auxiliary information thereof, and effectively transfer the knowledge learned by the known data set to the unknown data set, thereby realizing the classification of the unknown equipment data set.
Further, the step 1 of classifying the known type device and the unknown type device in the integrated energy system comprises:
step 1.1: n in the integrated energy systemknowThe time series of operating powers of the devices of known type form a known data setWherein DknowFor a known data set, pknow,iFor operating a power time series for a device of known type, yknow,iFor corresponding equipment tags of known type, YknowSet of device tags for known classes, QknowThe number of devices of known type;
step 1.2: n to be identified in the integrated energy systemXThe running power time sequence of the unknown type equipment forms an unknown data setWherein DXFor unknown data sets, pX,iTime series of operating powers for devices of unknown type, yknow,iFor corresponding device tags of unknown type, YknowFor unknown classes of device tag sets, QknowNumber of devices of known type, yX,iLabeling the corresponding unknown type device;
step 1.3: performing semantic dictionary learning on the running power time sequence of the known type equipment:
wherein L isknowFor known type device semantic dictionaries, HknowFor known type device semantic representation, | ·| non-woven calculationFIs the Frobenius norm,semantic dictionary regularization for known type devicesTerm, λ, controls the strength of the regularizing term, PknowRunning a power time series for a known type of device;
step 1.4: performing semantic dictionary learning on the running power time sequence of the unknown type equipment:
wherein,qjis yX,jRepresentation in semantic embedding space, ωijFor inputting a time sequence pX,iBelonging to the label yX,jProbability of (H)XFor semantic representation of unknown types of devices, LXFor a device semantic dictionary of unknown type, siFor the equipment running state, | LX-LknowI is for limiting LXAnd LknowRegular term of fitness, | hi-qjThe method comprises the following steps that I is a regular term used for limiting the similarity between the representation of the power time sequence of the unknown type equipment in the semantic embedding space and the representation of the label of the unknown type equipment in the semantic embedding space;
step 1.5: fixation of H in step 3.3 and step 3.4, respectivelyXAnd LXAnd performing alternate iterative solution, and finding corresponding classification in an embedding space according to the result:
in the above formula, the first and second carbon atoms are,the optimal solution is expressed semantically for the unknown type of device,representing an optimal solution for unknown type device semantics, | ·| non-calculationFIs Frobenius norm.
Furthermore, the step 2 of extracting the operation states of the devices in the integrated energy system according to the classification result utilizes the Bi-directional long-short term memory network-based device to extract the operation states, the used Bi-LSTM network comprises six layers, wherein the length of an input layer is the length of a time window t, the second layer is a convolutional layer for extracting features from signals, the third and fourth layers are Bi-LSTM, the fifth layer is a convolutional layer, the sixth layer is a fully-connected layer, and the whole network is trained by a time forward and backward Bi-directional propagation method.
Furthermore, the step 3 of dividing according to the extracted operation state includes the following steps:
step 3.1: and (3) dividing the running states of the electric equipment according to the classification result in the step 1:
wherein n isdThe number of the electric devices of the type d,the running state of the electric equipment at the time t;
step 3.2: dividing the operating states of the power generation equipment according to the classification result in the step 1:
wherein n isDGFor the number of operating modes of the power plant, sDGT is the running state of the power generation equipment at the moment t;
step 3.3: dividing the operating states of the energy storage equipment according to the classification result in the step 1:
wherein,for the mode of operation of the energy storage device,for the rated power of the energy storage device, ses,tThe operating state of the energy storage device at time t.
Further, step 4 is to establish a multi-modulus hidden markov model according to the division result, as follows:
in the above formula, the first and second carbon atoms are,operating state of the Q +1 st device at the initial moment, ptFor the operating power of the device at time t,for the operating state of the Q +1 st device at time t,operation status of the Q +1 th device at time t-1, N (μ)i,εi) To a desired value of muiStandard deviation of epsiloniThe normal distribution of (c),for the operating state of the 1 st device at time t,and the operation state of the 2 nd equipment at the moment t, wherein pi is the probability distribution of the initial state, A is a state transition matrix, and B is an observation matrix.
Further, the step 5 of performing parameter estimation on the established multi-modulus hidden markov model includes the following steps:
step 5.1: the multi-modulus hidden Markov model parameter estimation equation is established as follows:
in the above formula, θ*The observed multi-mode factor hidden Markov model parameter with the maximum probability of the time series of the operating power of the equipment, P is the operating power of the equipment, PtIs the running power of the equipment at the moment t, S is the running state of the equipment, StTheta is the operation state of the equipment at the time t, theta is a multi-modulus hidden Markov model parameter, | P1:TI is the running power time sequence of the equipment from the initial moment to the t moment;
step 5.2: introducing a forward auxiliary variableIs represented by the timetQ-type equipment operation power time sequence under operation state iGiven an initial parameter theta0,Expressed as:under the initial conditions of the process, the process is carried out,is shown as
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iOf joint probability of theta0For the multi-modulus hidden markov model initial parameters,the method comprises the following steps that (1) the operation state of q type equipment at an initial moment is obtained, and i is the operation state of the equipment at a t moment;
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iThe joint probability of (a) is determined,for the state transition probability of a q-type device,the q type device operating power for time t +1,the operating state of the q-type equipment at the moment t +1, the operating state of j at the moment t +1,operating power time sequence of q type equipment at t +1 moment and in operating state jA joint probability of (a);
step 5.4: introducing backward auxiliary variablesRepresenting a time series of operating powers of a device of type q observed at time t, operating state iGiven an initial parameter θ0,Expressed as:
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,the q type device operation state at the time t +1,the operating power of the q-type equipment at the moment t +1 is obtained;
step 5.6: for a given initial parameter θ0And observation sequenceComputing the slave state of a q-type deviceTransition to a StateProbability of (2)And presenting state at time tProbability of (2)
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,is the backward auxiliary variable at time t +1,for the backward auxiliary variable at time t,presenting state for time tThe probability of (a) of (b) being,slave status for q-type devicesTransition to a StateThe probability of (d);
step 5.7: recalculating the model parameters for the q hidden markov chains:
in the above formula, the first and second carbon atoms are,for the initial state probability estimate for a q-type device,for the q-type device state transition probability estimation,for an estimate of the expected values of the observation matrix for a q-type device,an estimated value of standard deviation of an observation matrix of the q-type equipment is obtained;
step 5.8: iterative computation of forward variables from new parameter loopsBackward variation Anduntil convergence.
Further, the step 6 of decoding the hidden state of the hidden markov model with the multi-modulus factor for the parameter estimation result to finally establish the comprehensive energy system decomposition model is to decode the hidden state of the hidden markov model with the multi-modulus factor by applying a Viterbi algorithm to the summarized power consumption sequence parameter estimation result to finally establish the comprehensive energy system decomposition model; the method comprises the following steps:
step 6.1: introducing a variable deltat(i) Representing all states s at time t1,s1,...,stTo an observation sequence p1,p2,...,pTMaximum value of probability:n is a device hidden layer state number, T is 1, 2.
Step 6.2: feeding in the product obtained in step 5.7And the output power time sequence P of the integrated energy system1:T={p1,p2,...,pT};
a computer storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of a zero sample learning based non-intrusive modeling method for an integrated energy system.
The invention has the following beneficial effects and advantages:
the invention provides a non-invasive modeling method of a comprehensive energy system based on zero sample learning, which can be used for carrying out non-invasive modeling on the comprehensive energy system containing unknown equipment. Firstly, extracting the running characteristics of known type equipment and unknown type equipment accessed to the comprehensive energy system through dictionary learning; further classifying equipment in the comprehensive energy system by using zero sample learning based on sparse coding; on the basis, the running state of the equipment is extracted by using a bidirectional long-short term memory network, and the modeling is carried out on the comprehensive energy system based on a multi-mode factor hidden Markov model. The method accurately identifies the unknown type equipment in the comprehensive energy system, and the established model effectively reduces the average error of the non-invasive modeling method, so that the precision and the accuracy of the model are obviously improved.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow chart of a non-intrusive modeling method of an integrated energy system based on zero sample learning according to the present invention;
FIG. 2 is a schematic diagram of the integrated energy system of the present invention;
FIG. 3 is a schematic diagram of the zero sample learning based integrated energy system classification of the present invention;
FIG. 4 is a block diagram of the bidirectional long short term memory network of the present invention;
FIG. 5 is a diagram of a multi-modal equipment operating status breakdown of the present invention;
FIG. 6 is a hidden Markov model with multi-modal factors in accordance with the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
The solution of some embodiments of the invention is described below with reference to fig. 1-6.
Example 1
The invention provides an embodiment, which is a non-intrusive modeling method of an integrated energy system based on zero sample learning, and the non-intrusive modeling method is realized by the non-intrusive modeling integrated energy system based on zero sample learning, as shown in fig. 1, and fig. 1 is a flow chart of the non-intrusive modeling method of the integrated energy system based on zero sample learning.
The invention specifically comprises the following steps:
the invention relates to equipment of known type in a non-intrusive modeling comprehensive energy system based on zero sample learning, which comprises: photovoltaic power generation equipment, wind power generation equipment, electric energy storage equipment, electric vehicles, rotating electrical machines, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.
The method comprises the following steps:
step 1.1: n in the integrated energy system shown in FIG. 2knowThe time series of operating powers of the devices of known type form a known data setWherein DknowFor a known data set, pknow,iFor operating a power time series for a device of known type, yknow,iFor corresponding equipment tags of known type, YknowSet of device tags for known classes, QknowThe number of devices of known type;
step 1.2: n to be identified in the integrated energy system shown in FIG. 2XThe running power time sequence of the unknown type equipment forms an unknown data setWherein DXFor unknown data sets, pX,iTime series of operating powers for devices of unknown type, yknow,iFor corresponding device tags of unknown type, YknowFor unknown classes of device tag sets, QknowNumber of devices of known type, yX,iAnd is labeled with the corresponding unknown type device.
Step 1.3: performing semantic dictionary learning on the running power time sequence of the known type equipment:
wherein L isknowFor known type device semantic dictionaries, HknowFor known type device semantic representation, | ·| non-woven calculationFIs the Frobenius norm,controlling the strength of the regularization term, P, for the known type device semantic dictionary regularization term, λknowA power time series is run for a known type of device.
Step 1.4: performing semantic dictionary learning on the running power time sequence of the unknown type equipment:
wherein,qjis yX,jRepresentation in semantic embedding space, ωijFor inputting a time sequence pX,iBelonging to the label yX,jProbability of (H)XFor semantic representation of unknown types of devices, LXFor a device semantic dictionary of unknown type, siThe equipment running state is set; l | |X-LknowI is for limiting LXAnd LknowRegular term of fitness, | hi-qjAnd | l is a regular term used for limiting the similarity between the representation of the power time sequence of the unknown type device in the semantic embedding space and the representation of the tag of the unknown type device in the semantic embedding space.
Step 1.5: fixation of H in step 3.3 and step 3.4, respectivelyXAnd LXAlternate iterative solution is performed, and the corresponding classification is found in the embedding space shown in fig. 3 according to the result:
in the above formula, the first and second carbon atoms are,the optimal solution is expressed semantically for the unknown type of device,representing an optimal solution for unknown type device semantics, | ·| non-calculationFIs Frobenius norm.
the operation state extraction is carried out by using a Bi-directional Long Short-Term Memory network (Bi-LSTM) device based on fig. 4, the Bi-LSTM network comprises six layers in total, wherein the length of an input layer is the length of a time window t, the second layer is a convolutional layer and is used for extracting characteristics from signals, the third layer and the fourth layer are Bi-LSTM, the fifth layer is a convolutional layer and the sixth layer is a full-connection layer, and the whole network is trained by a time forward and backward bidirectional propagation method.
And 3, dividing according to the extracted running state.
The method comprises the following steps of dividing the running states of electric equipment, power generation equipment and energy storage equipment in the comprehensive energy system according to the running states of the equipment extracted by Bi-LSTM, and converting the 0-1 switch running state division problem of the equipment into a multi-mode division problem of the same type of equipment according to the equipment type, as shown in FIG. 5, wherein FIG. 5 is a multi-mode equipment running state division diagram of the invention, and the method comprises the following steps:
step 3.1: and (3) dividing the running states of the electric equipment according to the classification result in the step 1:
wherein n isdThe number of the electric devices of the type d,and (3) the running state of the electric equipment at the time t.
Step 3.2: dividing the operating states of the power generation equipment according to the classification result in the step 1:
wherein n isDGFor the number of operating modes of the power plant, sDG,tIs the operating state of the power generating equipment at the time t.
Step 3.3: dividing the operating states of the energy storage equipment according to the classification result in the step 1:
wherein,for the mode of operation of the energy storage device,for the rated power of the energy storage device, ses,tThe operating state of the energy storage device at time t.
And 4, step 4: a multi-modulus hidden markov model is built according to the division result, as shown in fig. 6:
in the above formula, the first and second carbon atoms are,operating state of the Q +1 st device at the initial moment, ptFor the operating power of the device at time t,for the operating state of the Q +1 st device at time t,operation status of the Q +1 th device at time t-1, N (μ)i,εi) To a desired value of muiStandard deviation of epsiloniThe normal distribution of (c),for the operating state of the 1 st device at time t,and the operation state of the 2 nd equipment at the moment t, wherein pi is the probability distribution of the initial state, A is a state transition matrix, and B is an observation matrix.
And 5: the parameter estimation is carried out on the established multi-modulus factor hidden Markov model, and the method comprises the following steps:
step 5.1: the multi-modulus hidden Markov model parameter estimation equation is established as follows:
in the above formula, θ*The observed multi-mode factor hidden Markov model parameter with the maximum probability of the time series of the operating power of the equipment, P is the operating power of the equipment, PtIs the running power of the equipment at the moment t, S is the running state of the equipment, StTheta is the operation state of the equipment at the time t, theta is a multi-modulus hidden Markov model parameter, | P1:TAnd | is the running power time sequence of the equipment from the initial moment to the t moment.
Step 5.2: introducing a forward auxiliary variableRepresenting the operating power time series of a q-type device at time t in operating state iGiven an initial parameter theta0,Expressed as:under the initial conditions of the process, the process is carried out,is shown as
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iOf joint probability of theta0For the multi-modulus hidden markov model initial parameters,the device running state is the initial time q type device running state, and i is the t time device running state.
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iThe joint probability of (a) is determined,for the state transition probability of a q-type device,the q type device operating power for time t + 1,the operating state of the q-type equipment at the moment t +1, the operating state of j at the moment t +1,operating power time sequence of q type equipment at t +1 moment and in operating state jThe joint probability of (c).
Step 5.4: introducing backward auxiliary variablesRepresenting a time series of operating powers of a device of type q observed at time t, operating state iGiven an initial parameter θ0,Expressed as:
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,the q type device operation state at the time t + 1,and (5) operating power of the type q equipment at the moment t + 1.
Step 5.6: for a given initial parameter θ0And observation sequenceComputing the slave state of a q-type deviceTransition to a StateProbability of (2)And presenting state at time tProbability of (2)
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,is the backward auxiliary variable at time t + 1,for the backward auxiliary variable at time t,presenting state for time tThe probability of (a) of (b) being,slave status for q-type devicesTransition to a StateThe probability of (c).
Step 5.7: recalculating the model parameters for the q hidden markov chains:
in the above formula, the first and second carbon atoms are,for the initial state probability estimate for a q-type device,for the q-type device state transition probability estimation,for an estimate of the expected values of the observation matrix for a q-type device,and (4) observing the estimated value of the standard deviation of the matrix for the q-type equipment.
Step 5.8: iterative computation of forward variables from new parameter loopsBackward variation Anduntil convergence.
Step 6: and decoding the hidden state of the multi-mode factor hidden Markov model by applying a Viterbi algorithm to the summarized power consumption sequence parameter estimation result, and finally establishing a comprehensive energy system decomposition model.
The method specifically comprises the following steps:
step 6.1: introducing a variable deltat(i) Representing all states s at time t1,s1,...,stTo an observation sequence p1,p2,...,pTMaximum value of probability:n is the device hidden layer state number, and T is 1, 2.
Step 6.2: feeding in the product obtained in step 5.7And the output power time sequence P of the integrated energy system1:T={p1,p2,...,pT};
the embodiment achieves the following technical effects:
(1) the unknown equipment type accessed into the comprehensive energy system can be accurately identified;
(2) the running states of the electric equipment, the energy storage equipment and the power generation equipment in the comprehensive energy system can be accurately extracted, and the running states are divided according to the equipment types;
(3) based on equipment type identification and equipment running state extraction, the running state of each equipment in the comprehensive energy system and the overall energy output condition of the comprehensive energy system can be accurately reflected by adopting a factor hidden Markov model.
(4) According to the method, the operation states of the equipment in the comprehensive energy system are divided, so that the calculation complexity and the calculation time of the comprehensive energy system modeling are effectively reduced, and meanwhile, the unknown equipment is classified by adopting zero-sample learning, so that the precision and the accuracy of the model are improved.
Example 2
Based on the same inventive concept, an embodiment of the present invention further provides a computer storage medium, where a computer program is stored on the computer storage medium, and when the computer program is executed by a processor, the computer program implements the steps of the non-invasive modeling method for an integrated energy system based on zero sample learning according to embodiment 1.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (10)
1. A non-invasive modeling method of a comprehensive energy system based on zero sample learning is characterized by comprising the following steps: the method comprises the following steps:
step 1, classifying known type equipment and unknown type equipment in the comprehensive energy system;
step 2, extracting the running state of each device in the comprehensive energy system according to the classification result;
step 3, dividing according to the extracted running state;
and 4, step 4: establishing a multi-mode factor hidden Markov model according to the division result;
and 5: performing parameter estimation on the established multi-mode factor hidden Markov model;
step 6: and decoding the hidden state of the multi-mode factor hidden Markov model according to the parameter estimation result, and finally establishing a comprehensive energy system decomposition model.
2. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: the known types of devices in said integrated energy system comprise: photovoltaic power generation equipment, wind power generation equipment, electric energy storage equipment, electric vehicles, rotating electrical machines, electric boilers, micro gas turbines, electric heaters, electric ovens, refrigerators, washing machines and lighting equipment.
3. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: the step of classifying the known type equipment and the unknown type equipment is to classify the equipment in the self-energy source by utilizing zero sample learning based on sparse coding, train a classification model by utilizing a known data set and auxiliary information of a comprehensive energy system, and effectively transfer the knowledge learned by the known data set to the unknown data set, thereby realizing the classification of the unknown equipment data set.
4. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: the step 1 of classifying the known type equipment and the unknown type equipment in the comprehensive energy system comprises the following steps:
step 1.1: n in the integrated energy systemknowThe time series of operating powers of the devices of known type form a known data setWherein DknowFor a known data set, pknow,iFor operating a power time series for a device of known type, yknow,iFor corresponding equipment tags of known type, YknowSet of device tags for known classes, QknowThe number of devices of known type;
step 1.2: n to be identified in the integrated energy systemXThe running power time sequence of the unknown type equipment forms an unknown data setWherein DXFor unknown data sets, pX,iTime series of operating powers for devices of unknown type, yknow,iFor corresponding device tags of unknown type, YknowFor unknown classes of device tag sets, QknowNumber of devices of known type, yX,iLabeling the corresponding unknown type device;
step 1.3: performing semantic dictionary learning on the running power time sequence of the known type equipment:
wherein L isknowFor known type device semantic dictionaries, HknowFor known type device semantic representation, | ·| non-woven calculationFIs the Frobenius norm,controlling the strength of the regularization term, P, for the known type device semantic dictionary regularization term, λknowRunning a power time series for a known type of device;
step 1.4: performing semantic dictionary learning on the running power time sequence of the unknown type equipment:
wherein,qjis yX,jRepresentation in semantic embedding space, ωijFor inputting a time sequence pX,iBelonging to the label yX,jProbability of (H)XFor semantic representation of unknown types of devices, LXFor a device semantic dictionary of unknown type, siFor the equipment running state, | LX-LknowI is for limiting LXAnd LknowRegular term of fitness, | hi-qjThe method comprises the following steps that I is a regular term used for limiting the similarity between the representation of the power time sequence of the unknown type equipment in the semantic embedding space and the representation of the label of the unknown type equipment in the semantic embedding space;
step 1.5: fixation of H in step 3.3 and step 3.4, respectivelyXAnd LXAnd performing alternate iterative solution, and finding corresponding classification in an embedding space according to the result:
5. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: and 2, extracting the running state of each device in the comprehensive energy system according to the classification result, namely extracting the running state by using the device based on the bidirectional long-short term memory network, wherein the used Bi-LSTM network comprises six layers, the length of an input layer is the length of a time window t, the second layer is a convolutional layer and is used for extracting characteristics from signals, the third layer and the fourth layer are Bi-LSTM, the fifth layer is a convolutional layer, the sixth layer is a full-connection layer, and the whole network is trained by a time forward-backward bidirectional transmission method.
6. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: the step 3 of dividing according to the extracted running state comprises the following steps:
step 3.1: and (3) dividing the running states of the electric equipment according to the classification result in the step 1:
wherein n isdThe number of the electric devices of the type d,the running state of the electric equipment at the time t;
step 3.2: dividing the operating states of the power generation equipment according to the classification result in the step 1:
wherein n isDGFor the number of operating modes of the power plant, sDG,tThe operation state of the power generation equipment at the time t is shown;
step 3.3: dividing the operating states of the energy storage equipment according to the classification result in the step 1:
7. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: step 4, establishing a multi-modulus factor hidden Markov model according to the division result, as follows:
in the above formula, the first and second carbon atoms are,operating state of the Q +1 st device at the initial moment, ptFor the operating power of the device at time t,for the operating state of the Q +1 st device at time t,operation status of the Q +1 th device at time t-1, N (μ)i,εi) To a desired value of muiStandard deviation of epsiloniThe normal distribution of (c),for the operating state of the 1 st device at time t,and the operation state of the 2 nd equipment at the moment t, wherein pi is the probability distribution of the initial state, A is a state transition matrix, and B is an observation matrix.
8. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: and 5, performing parameter estimation on the established multi-modulus factor hidden Markov model, comprising the following steps of:
step 5.1: the multi-modulus hidden Markov model parameter estimation equation is established as follows:
in the above formula, θ*The observed multi-mode factor hidden Markov model parameter with the maximum probability of the time series of the operating power of the equipment, P is the operating power of the equipment, PtIs the running power of the equipment at the moment t, S is the running state of the equipment, StTheta is the operation state of the equipment at the time t, theta is a multi-modulus hidden Markov model parameter, | P1:TI is the running power time sequence of the equipment from the initial moment to the t moment;
step 5.2: introducing a forward auxiliary variableRepresenting the operating power time series of a q-type device at time t in operating state iGiven an initial parameter theta0,Expressed as:under the initial conditions of the process, the process is carried out,is shown as
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iOf joint probability of theta0For the multi-modulus hidden markov model initial parameters,the method comprises the following steps that (1) the operation state of q type equipment at an initial moment is obtained, and i is the operation state of the equipment at a t moment;
In the above formula, the first and second carbon atoms are,for q type equipment operation power time sequence under t time and operation state iThe joint probability of (a) is determined,for the state transition probability of a q-type device,the q type device operating power for time t +1,the operating state of the q-type equipment at the moment t +1, the operating state of j at the moment t +1,operating power time sequence of q type equipment at t +1 moment and in operating state jA joint probability of (a);
step 5.4: introducing backward auxiliary variablesRepresenting a time series of operating powers of a device of type q observed at time t, operating state iGiven an initial parameter θ0,Expressed as:
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,the q type device operation state at the time t +1,the operating power of the q-type equipment at the moment t +1 is obtained;
step 5.6: for a given initial parameter θ0And observation sequenceComputing the slave state of a q-type deviceTransition to a StateProbability of (2)And presenting state at time tProbability of (2)
In the above formula, the first and second carbon atoms are,for the state transition probability of a q-type device,is the backward auxiliary variable at time t +1,for the backward auxiliary variable at time t,presenting state for time tThe probability of (a) of (b) being,slave status for q-type devicesTransition to a StateThe probability of (d);
step 5.7: recalculating the model parameters for the q hidden markov chains:
in the above formula, the first and second carbon atoms are,for the initial state probability estimate for a q-type device,for the q-type device state transition probability estimation,for an estimate of the expected values of the observation matrix for a q-type device,an estimated value of standard deviation of an observation matrix of the q-type equipment is obtained;
9. The non-invasive modeling method of the integrated energy system based on the zero sample learning as claimed in claim 1, wherein: step 6, decoding the hidden state of the multi-mode factor hidden Markov model of the parameter estimation result, and finally establishing a comprehensive energy system decomposition model, namely decoding the hidden state of the multi-mode factor hidden Markov model by applying a Viterbi algorithm to the summarized power consumption sequence parameter estimation result, and finally establishing the comprehensive energy system decomposition model; the method comprises the following steps:
step 6.1: introducing a variable deltat(i) Representing all states s at time t1,s1,...,stTo an observation sequence p1,p2,...,pTMaximum value of probability:n is a device hidden layer state number, T is 1, 2.
Step 6.2: feeding in the product obtained in step 5.7And the output power time sequence P of the integrated energy system1:T={p1,p2,...,pT};
10. A computer storage medium, characterized by: the computer storage medium has stored thereon a computer program that, when executed by a processor, performs the steps of a method for non-intrusive modeling of an integrated energy system based on zero sample learning of claims 1-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111495935.8A CN114386314A (en) | 2021-12-09 | 2021-12-09 | Non-invasive modeling method of comprehensive energy system based on zero sample learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111495935.8A CN114386314A (en) | 2021-12-09 | 2021-12-09 | Non-invasive modeling method of comprehensive energy system based on zero sample learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114386314A true CN114386314A (en) | 2022-04-22 |
Family
ID=81196719
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111495935.8A Pending CN114386314A (en) | 2021-12-09 | 2021-12-09 | Non-invasive modeling method of comprehensive energy system based on zero sample learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114386314A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115146876A (en) * | 2022-09-05 | 2022-10-04 | 常州威图维亚电机科技有限公司 | Electric vehicle power optimal control method based on energy model |
CN116595488A (en) * | 2023-07-19 | 2023-08-15 | 青岛鼎信通讯股份有限公司 | Non-invasive load identification method based on intelligent ammeter |
CN117728583A (en) * | 2023-12-27 | 2024-03-19 | 中节能甘肃武威太阳能发电有限公司 | Distributed photovoltaic cluster energy control monitoring system based on transfer learning |
-
2021
- 2021-12-09 CN CN202111495935.8A patent/CN114386314A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115146876A (en) * | 2022-09-05 | 2022-10-04 | 常州威图维亚电机科技有限公司 | Electric vehicle power optimal control method based on energy model |
CN115146876B (en) * | 2022-09-05 | 2022-11-11 | 常州威图维亚电机科技有限公司 | Electric vehicle power optimal control method based on energy model |
CN116595488A (en) * | 2023-07-19 | 2023-08-15 | 青岛鼎信通讯股份有限公司 | Non-invasive load identification method based on intelligent ammeter |
CN116595488B (en) * | 2023-07-19 | 2023-11-14 | 青岛鼎信通讯股份有限公司 | Non-invasive load identification method based on intelligent ammeter |
CN117728583A (en) * | 2023-12-27 | 2024-03-19 | 中节能甘肃武威太阳能发电有限公司 | Distributed photovoltaic cluster energy control monitoring system based on transfer learning |
CN117728583B (en) * | 2023-12-27 | 2024-05-31 | 中节能甘肃武威太阳能发电有限公司 | Distributed photovoltaic cluster energy control monitoring system based on transfer learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhou et al. | One fits all: Power general time series analysis by pretrained lm | |
CN114386314A (en) | Non-invasive modeling method of comprehensive energy system based on zero sample learning | |
Chang et al. | Chinese named entity recognition method based on BERT | |
CN112784965A (en) | Large-scale multi-element time series data abnormity detection method oriented to cloud environment | |
Abdel-Galil et al. | Disturbance classification using hidden Markov models and vector quantization | |
Sun et al. | TEST: Text prototype aligned embedding to activate LLM's ability for time series | |
Liu et al. | Dynamic knowledge graph reasoning based on deep reinforcement learning | |
Ma et al. | A survey on time-series pre-trained models | |
CN116468298A (en) | GPT network model-based automobile technology planning and decision-making method and system | |
CN113868432B (en) | Automatic knowledge graph construction method and system for iron and steel manufacturing enterprises | |
CN113934887A (en) | No-proposal time sequence language positioning method based on semantic decoupling | |
Liao et al. | A machine learning approach to case adaptation | |
CN116108143B (en) | Digital economic monitoring method and system based on block chain technology | |
CN112835585A (en) | Program understanding method and system based on abstract syntax tree | |
Eldele et al. | Label-efficient time series representation learning: A review | |
Lin et al. | Research on Named Entity Recognition Method of Metro On‐Board Equipment Based on Multiheaded Self‐Attention Mechanism and CNN‐BiLSTM‐CRF | |
CN118335190A (en) | Method and system for generating amino acid sequence of protein with specific functions and properties by using deep learning technology | |
Chen et al. | Knowledge-aware textual entailment with graph attention network | |
Wang et al. | Out-of-distributed semantic pruning for robust semi-supervised learning | |
Zhang et al. | Cross-Supervised multisource prototypical network: A novel domain adaptation method for multi-source few-shot fault diagnosis | |
Zhang | Machine Learning and Visual Perception | |
CN114596473A (en) | Network embedding pre-training method based on graph neural network hierarchical loss function | |
Gao et al. | Imprecise deep Forest for partial label learning | |
Liu et al. | Deep learning based appliance identification | |
Li et al. | Graph Intelligence with Large Language Models and Prompt Learning |
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 |