CN115730710B - Electric vehicle daily charging demand curve prediction method based on attention mechanism - Google Patents
Electric vehicle daily charging demand curve prediction method based on attention mechanism Download PDFInfo
- Publication number
- CN115730710B CN115730710B CN202211415516.3A CN202211415516A CN115730710B CN 115730710 B CN115730710 B CN 115730710B CN 202211415516 A CN202211415516 A CN 202211415516A CN 115730710 B CN115730710 B CN 115730710B
- Authority
- CN
- China
- Prior art keywords
- electric automobile
- state
- layer
- probability
- attention
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000007246 mechanism Effects 0.000 title claims abstract description 28
- 238000004088 simulation Methods 0.000 claims abstract description 58
- 238000013528 artificial neural network Methods 0.000 claims abstract description 43
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 11
- 238000010606 normalization Methods 0.000 claims description 72
- 230000007704 transition Effects 0.000 claims description 50
- 239000011159 matrix material Substances 0.000 claims description 41
- 238000005070 sampling Methods 0.000 claims description 30
- 238000004364 calculation method Methods 0.000 claims description 26
- 230000009466 transformation Effects 0.000 claims description 23
- 238000005315 distribution function Methods 0.000 claims description 16
- 238000012423 maintenance Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 description 5
- 238000000342 Monte Carlo simulation Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 230000003920 cognitive function Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000013486 operation strategy Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000528 statistical test Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/60—Other road transportation technologies with climate change mitigation effect
- Y02T10/70—Energy storage systems for electromobility, e.g. batteries
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/60—Other road transportation technologies with climate change mitigation effect
- Y02T10/7072—Electromobility specific charging systems or methods for batteries, ultracapacitors, supercapacitors or double-layer capacitors
Landscapes
- Charge And Discharge Circuits For Batteries Or The Like (AREA)
Abstract
The invention provides an electric vehicle daily charging demand curve prediction method based on an attention mechanism, which relates to the technical field of electric vehicle charging demand power prediction and comprises the following steps: acquiring initial data of the electric vehicle in each time slot in a day, acquiring a historical load curve of the electric vehicle in a day according to the initial data, pre-simulating a daily charging demand simulation curve of the electric vehicle through a Monte Carlo algorithm and a preset Markov model, combining the daily charging demand simulation curve with the historical load curve, and finally acquiring a daily charging demand prediction curve of the electric vehicle by using the attention mechanism of a Transformer neural network; the method can analyze potential correlation between global randomness information and local deterministic information of the daily charging power of the electric automobile, has certain nonlinear fitting capacity, and the daily charging demand prediction curve of the electric automobile obtained by the method has the characteristics of high prediction precision and strong robustness and has strong prediction performance.
Description
Technical Field
The invention relates to the technical field of electric vehicle charging demand power prediction, in particular to an electric vehicle daily charging demand curve prediction method based on an attention mechanism.
Background
As environmental and energy problems are increasingly prominent, domestic electric vehicles are rapidly developed in recent years as representatives of clean energy vehicles, and charging loads of the electric vehicles also show statistical characteristics. With the continuous improvement of power batteries and vehicle technologies, the large-scale application of electric automobiles is increasing. Under the background, the characteristic analysis and prediction are carried out on the charging demand load of the electric automobile, so that on one hand, the economic operation and energy management of the charging station of the electric automobile are facilitated, and reference data are provided for planning and construction of urban infrastructure and the like; on the other hand, the method is favorable for optimal distribution of a power system and economic dispatching of a power grid, and has profound significance for electric market trading, optimal combination research of a generator set and the like.
The monte carlo method in computer simulation is also called a random sampling technique or a statistical test method, and the most important characteristic of the method is that the method is a method based on a probability statistical theory. Along with the development of science and technology and the invention of an electronic computer, the Monte Carlo method has been widely applied in various fields by describing the physical development characteristics and the advantages of the physical experiment process.
Meanwhile, the transducer neural network is widely applied to the field of power prediction due to the strong modeling capability of the transducer neural network on time series data, supports parallelism, trains faster and has strong long-term dependence modeling capability, which is greatly helpful for daily charge demand prediction, and meanwhile, the attention score (attention score) of the transducer can provide a certain interpretation. The distribution of current predictions over historical value attention can be seen by visualizing the actionscore. The whole network structure of the transducer mainly comprises two parts of an Encoder and a Decoder, and most of sequence models (Encoder-Decoders) are based on CNN and RNN neural networks before the transducer neural network is not produced.
Attention is an indispensable complex cognitive function of humans, which refers to the selective ability to concentrate on certain information and ignore other information. The attention mechanism utilizes the human brain to improve the ability of neural networks to process information. When the neural network processes a large amount of input information, the Attention mechanism allows the network to select only some key information as input, while the Transformer neural network is based on the Attention mechanism, the Attention mechanism is better than the CNN and RNN neural networks in certain aspects, the Attention mechanism can solve the long-distance dependence problem of the RNN and the variants thereof, namely, the Attention mechanism can have better memory and can memorize longer-distance information, and in addition, the most important is that the Attention mechanism supports parallelized calculation. The transducer model is based entirely on the Attention mechanism, which completely discards the structure of CNN and RNN neural networks.
The prior art discloses a method for predicting the large-scale charging demand of an electric automobile based on a Monte Carlo simulation method, which comprises the following steps: respectively establishing a slow charge demand prediction model and a fast charge prediction model of the electric automobile to obtain a slow charge power expected and a fast charge power expected of a single electric automobile in one day, wherein the Monte Carlo simulation method has certain randomness, so that repeated tests are needed to be carried out and then average value is obtained to ensure the accuracy of prediction, and further a charging demand characteristic curve of the large-scale electric automobile in a planning area is obtained; the method in the prior art does not add an attention mechanism, and only predicts by a base Yu Mengte Carlo algorithm; the Monte Carlo algorithm is based on the operation characteristic analysis of the electric automobile, even if a plurality of scenes are set, the randomness of the charging of the electric automobile is difficult to accurately and completely reflect, and the utilization of deterministic historical load information is lacking, so that the prediction result is a coarse-scale expected prediction fundamentally, a certain basis and reference can be provided for a power grid capacity-increasing and transformation plan, peak regulation capacity construction, long-term scheduling operation strategy and the like, the random charging load behaviors of the large-scale electric automobile in different states on a certain day can not be simulated under the condition of smaller time resolution, and particularly when a simulated system has some special external influence factors (major gathering, natural disasters, power failure and the like), the prediction result deviates from the actual condition.
Disclosure of Invention
The invention provides an electric vehicle daily charging demand curve prediction method based on an attention mechanism, which can obtain stronger prediction performance in order to overcome the defects of low prediction precision and poor robustness of the electric vehicle daily charging power demand curve in the prior art.
In order to solve the technical problems, the technical scheme of the invention is as follows:
an electric automobile daily charging demand curve prediction method based on an attention mechanism comprises the following steps:
s1: acquiring initial data of the electric automobile in each time slot in a day, and acquiring a historical load curve of the electric automobile in a day according to the acquired initial data of the electric automobile;
s2: setting basic parameters of an electric automobile, randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation, and obtaining a daily charging demand simulation curve of the electric automobile;
s3: encoding the daily charging demand simulation curve by using a preset global randomness information encoder, encoding the historical load curve by using a preset local deterministic information encoder, and setting a position code;
S4: fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a preset transducer encoder together with preset auxiliary information to obtain a coding result;
s5: inputting the daily charge demand simulation curve and the encoding result into a preset first converter decoder for decoding to obtain a global information decoding result, and inputting the historical load curve and the encoding result into a preset second converter decoder for decoding to obtain a local information decoding result;
s6: and fusing the global information decoding result and the local information decoding result to obtain a daily charging demand prediction curve of the electric automobile.
Preferably, the initial data of the electric vehicle in step S1 specifically includes:
the initial data of the electric automobile comprises: state data of the electric automobile, state of charge (SOC) data of a battery of the electric automobile, initial charging time data of the electric automobile and daily driving mileage data of the electric automobile;
the states of the electric automobile include: state 1: normal state of charge, state 2: state of fast charge, state 3: running state and state 4: and (5) a parking state.
Preferably, in step S2, basic parameters of the electric vehicle are set, the initial data of the electric vehicle obtained is randomly sampled by adopting a monte carlo algorithm, sampling data of a plurality of time slots of the electric vehicle are obtained, the sampling data of the plurality of time slots of the electric vehicle are input into a preset markov model for calculation, and a daily charging demand simulation curve of the electric vehicle is obtained, and the specific method is as follows:
The basic parameters of the electric automobile comprise: the number, type, maximum battery capacity and state of charge (SOC) constraints of the batteries of the electric vehicle;
s2.1: randomly sampling the obtained initial data of the electric automobile to obtain sampling data of a plurality of time slots of the electric automobile;
s2.2: inputting sampling data of a plurality of time slots of the electric automobile into a preset Markov model, and calculating a Markov transition probability matrix of the electric automobile;
s2.3: calculating a time slot state probability distribution function of the electric automobile according to the Markov transition probability matrix of the electric automobile;
s2.4: and obtaining a daily charging demand simulation curve of the electric automobile according to the time slot state probability distribution function of the electric automobile.
Preferably, in the step S2.2, the markov transition probability matrix of the electric automobile is specifically:
wherein P represents a Markov transition probability matrix of the electric automobile, and P 11 Representing probability of maintaining state 1 of electric automobile in next time slot, P 12 Representing probability of electric automobile transition from state 1 to state 2, P 13 Representing probability of electric automobile transition from state 1 to state 3, P 14 The probability of the electric automobile transitioning from state 1 to state 4 is represented; p (P) 22 Representing probability of maintaining state 2 of electric automobile in next time slot, P 21 Indicating the transition of the electric automobile from state 2 to stateProbability of 1, P 23 Representing probability of electric automobile transition from state 2 to state 3, P 24 The probability of the electric automobile transitioning from state 2 to state 4 is represented; p (P) 33 Representing probability of maintaining state 3 of electric automobile in next time slot, P 31 Representing probability of transition of electric automobile from state 3 to state 1, P 32 Representing probability of transition of electric automobile from state 3 to state 2, P 44 The probability of the electric automobile transitioning from state 3 to state 4 is represented; p (P) 44 Representing probability of electric automobile maintenance state 4 in next time slot, P 41 Representing probability of transition of electric automobile from state 4 to state 1, P 42 Representing probability of electric automobile transition from state 4 to state 2, P 43 The probability of the electric vehicle transitioning from state 4 to state 3 is shown.
Preferably, in the step S2.3, the probability distribution function of the time slot state of the electric automobile is specifically:
the electric automobile state probability distribution function of each time slot in one day is specifically:
wherein ,representing the probability of normal charging state of the electric automobile at the moment t, < >>Probability of representing fast charge state of electric vehicle at time t,/->Representing the probability of the running state of the electric vehicle at the time t, < + >>And the probability of the parking state of the electric automobile at the time t is shown.
Preferably, in step S2.4, the specific method for obtaining the daily charging demand simulation curve of the electric vehicle according to the time slot state probability distribution function of the electric vehicle is as follows:
obtaining a daily charging demand simulation curve of the electric automobile according to the following formula:
wherein V represents a daily charging demand analog value of the electric automobile, and P n Indicating the power required by the normal charge state of the electric automobile, P f The power required by the electric automobile in the quick charge state is represented, and n represents the number of the electric automobiles in the sampled data.
Preferably, the position coding in the step S3 is specifically:
the position code is used for marking the sequence of the input transducer encoder in the time dimension, and the calculation formula is as follows:
wherein ,PE(pos,2j) Position coding representing even time dimension, PE (pos,2j+1) Position code representing odd time dimension, pos representing the position of the current slot in the day, d model Representing the overall dimension of the position code; 2j represents even time dimension, 2j+1 represents odd time dimension, satisfying 2 j.ltoreq.d model ，2j+1≤d model 。
Preferably, in the step S4, the encoded daily charging demand simulation curve, the encoded historical load curve and the encoded position code are fused, and are input into a preset transform encoder together with preset auxiliary information, so as to obtain an encoding result, and the specific method is as follows:
The preset auxiliary information comprises weather information and temperature information;
the preset transducer encoder comprises a first multi-head self-attention layer, a first normalization layer, a first feedforward neural network layer and a second normalization layer which are sequentially connected;
the input end of the first multi-head self-attention layer and the input end of the first normalization layer form residual connection, and the input end of the first feedforward neural network layer and the input end of the second normalization layer form residual connection;
fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a first multi-head self-attention layer together with preset auxiliary information, wherein the input of the first multi-head self-attention layer is marked as a Z vector;
mapping the Z vector to Q, K and V matrixes through h different linear transforms respectively, and obtaining h attention matrixes Z through self-attention calculation and linear transforms respectively 1 ；
All attention matrix Z 1 Splicing to obtain a spliced matrix, and combining the spliced matrix with a preset weight matrix W 0 Multiplying to obtain a splice matrix Z containing attention information 2 ；
Splice matrix Z to contain attention information 2 Calculating and outputting through the first normalization layer, the first feedforward neural network layer and the second normalization layer to obtain a coding result, and marking the coding result as Z 3 Vector.
Preferably, in step S5, the daily charge demand simulation curve and the encoding result are input into a preset first converter decoder to decode, so as to obtain a global information decoding result, and the historical load curve and the encoding result are input into a preset second converter decoder to decode, so as to obtain a local information decoding result, and the specific method is as follows:
the first and second transducers are arranged in parallel;
the first transducer decoder comprises a second multi-head self-attention layer, a third normalization layer, a first encoder-decoder attention layer, a fourth normalization layer, a second feedforward neural network layer, a fifth normalization layer and a first linear transformation layer which are sequentially connected;
the input end of the second multi-head self-attention layer is in residual connection with the input end of the third normalization layer, the input end of the first encoder-decoder attention layer is in residual connection with the input end of the fourth normalization layer, and the input end of the second feedforward neural network layer is in residual connection with the input end of the fifth normalization layer;
the second transducer decoder comprises a third multi-head self-attention layer, a sixth normalization layer, a second encoder-decoder attention layer, a seventh normalization layer, a third feedforward neural network layer, an eighth normalization layer and a second linear transformation layer which are sequentially connected;
The input end of the third multi-head self-attention layer is in residual connection with the input end of the sixth normalization layer, the input end of the second encoder-decoder attention layer is in residual connection with the input end of the seventh normalization layer, and the input end of the third feedforward neural network layer is in residual connection with the input end of the eighth normalization layer;
taking a daily charge demand simulation curve as input data of a second multi-head self-attention layer, taking a coding result of a transducer coder and output data of a third normalization layer as input data of a first coder-decoder attention layer, performing linear transformation calculation on output data of a fifth normalization layer through a first linear transformation layer to obtain a global information decoding result, and marking the global information decoding result as Z 4 Vector;
taking the historical load curve as input data of a third multi-head self-attention layer, taking the coding result of a transducer coder and the output data of a sixth normalization layer as input data of a second coder-decoder attention layer, carrying out linear transformation calculation on the output data of an eighth normalization layer through a second linear transformation layer to obtain a local information decoding result, and marking the local information decoding result as Z 5 Vector.
Preferably, the first multi-head self-attention layer, the second multi-head self-attention layer and the third multi-head self-attention layer are identical in structure;
The first feedforward neural network layer, the second feedforward neural network layer and the third feedforward neural network layer have the same structure and all comprise two full-connection layers.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
the invention provides an electric vehicle daily charging demand curve prediction method based on an attention mechanism, which comprises the steps of firstly acquiring electric vehicle initial data of each time slot in a day, acquiring a daily historical load curve of the electric vehicle according to the acquired electric vehicle initial data, pre-simulating a daily charging demand simulation curve of the electric vehicle through a Monte Carlo algorithm and a preset Markov model, combining the daily charging demand simulation curve with the historical load curve, and finally acquiring the daily charging demand prediction curve of the electric vehicle by using the attention mechanism of a Transformer neural network;
the invention fully focuses on the potential correlation between the global randomness information and the local deterministic information of the daily charging power of the electric automobile, considers the inherent difference between different information, adopts a double-decoder structure, can ensure that decoders of specific tasks keep enough nonlinear fitting capacity and focuses on the prediction characteristics of the respective tasks, so that the finally obtained daily charging demand prediction curve of the electric automobile has the characteristics of high prediction precision and strong robustness, and can obtain stronger prediction performance.
Drawings
Fig. 1 is a flowchart of a method for predicting a daily charging demand curve of an electric vehicle based on an attention mechanism according to embodiment 1.
Fig. 2 is a flowchart of the computation of the transducer encoder according to embodiment 2.
Fig. 3 is a flowchart of the first multi-headed self-care layer calculation provided in embodiment 2.
Fig. 4 is a flowchart of the calculation of the first and second transducers according to embodiment 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, the embodiment provides a method for predicting a daily charging demand curve of an electric vehicle based on an attention mechanism, which includes the following steps:
s1: acquiring initial data of the electric automobile in each time slot in a day, and acquiring a historical load curve of the electric automobile in a day according to the acquired initial data of the electric automobile;
S2: setting basic parameters of an electric automobile, randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation, and obtaining a daily charging demand simulation curve of the electric automobile;
s3: encoding the daily charging demand simulation curve by using a preset global randomness information encoder, encoding the historical load curve by using a preset local deterministic information encoder, and setting a position code;
s4: fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a preset transducer encoder together with preset auxiliary information to obtain a coding result;
s5: inputting the daily charge demand simulation curve and the encoding result into a preset first converter decoder for decoding to obtain a global information decoding result, and inputting the historical load curve and the encoding result into a preset second converter decoder for decoding to obtain a local information decoding result;
s6: and fusing the global information decoding result and the local information decoding result to obtain a daily charging demand prediction curve of the electric automobile.
In the specific implementation process, firstly, electric vehicle initial data of each time slot in a day are obtained, and a historical load curve of the electric vehicle in the day is obtained according to the obtained electric vehicle initial data; setting basic parameters of an electric automobile, randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation, and obtaining a daily charging demand simulation curve of the electric automobile; then, a preset global randomness information encoder is used for encoding the daily charging demand simulation curve, a preset local deterministic information encoder is used for encoding the historical load curve, and position encoding is set; fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a preset transducer encoder together with preset auxiliary information to obtain a coding result; inputting the daily charge demand simulation curve and the encoding result into a preset first converter decoder for decoding to obtain a global information decoding result, and inputting the historical load curve and the encoding result into a preset second converter decoder for decoding to obtain a local information decoding result; finally, the global information decoding result and the local information decoding result are fused to obtain an electric vehicle daily charging demand prediction curve, and the obtained electric vehicle daily charging demand prediction curve is used for planning the optimal distribution of a power system and the economic dispatch of a power grid;
The invention fully focuses on the potential correlation between the global randomness information and the local deterministic information of the daily charging power of the electric automobile, considers the inherent difference between different information, adopts a double-decoder structure, can ensure that decoders of specific tasks keep enough nonlinear fitting capacity and focuses on the prediction characteristics of the respective tasks, so that the finally obtained daily charging demand prediction curve of the electric automobile has the characteristics of high prediction precision and strong robustness, and can obtain stronger prediction performance.
Example 2
The embodiment provides an electric vehicle daily charging demand curve prediction method based on an attention mechanism, which comprises the following steps:
s1: acquiring initial data of the electric automobile in each time slot in a day, and acquiring a historical load curve of the electric automobile in a day according to the acquired initial data of the electric automobile;
the initial data of the electric automobile comprises: state data of the electric automobile, state of charge (SOC) data of a battery of the electric automobile, initial charging time data of the electric automobile and daily driving mileage data of the electric automobile;
the states of the electric automobile include: state 1: normal state of charge, state 2: state of fast charge, state 3: running state and state 4: a parking state;
S2: setting basic parameters of an electric automobile, randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation to obtain a daily charging demand simulation curve of the electric automobile, wherein the method specifically comprises the following steps:
the basic parameters of the electric automobile comprise: number, type, maximum battery capacity, and state of charge, SOC, constraints for the battery of an electric vehicle
S2.1: randomly sampling the obtained initial data of the electric automobile to obtain sampling data of a plurality of time slots of the electric automobile;
s2.2: inputting sampling data of a plurality of time slots of the electric automobile into a preset Markov model, and calculating a Markov transition probability matrix of the electric automobile, wherein the Markov transition probability matrix of the electric automobile is specifically as follows:
wherein P represents a Markov transition probability matrix of the electric automobile, and P 11 Representing probability of maintaining state 1 of electric automobile in next time slot, P 12 Representing probability of electric automobile transition from state 1 to state 2, P 13 Representing probability of electric automobile transition from state 1 to state 3, P 14 Indicating the transition of the electric automobile from state 1 to the state Probability of state 4; p (P) 22 Representing probability of maintaining state 2 of electric automobile in next time slot, P 21 Representing probability of transition of electric automobile from state 2 to state 1, P 23 Representing probability of electric automobile transition from state 2 to state 3, P 24 The probability of the electric automobile transitioning from state 2 to state 4 is represented; p (P) 33 Representing probability of maintaining state 3 of electric automobile in next time slot, P 31 Representing probability of transition of electric automobile from state 3 to state 1, P 32 Representing probability of transition of electric automobile from state 3 to state 2, P 34 The probability of the electric automobile transitioning from state 3 to state 4 is represented; p (P) 44 Representing probability of electric automobile maintenance state 4 in next time slot, P 41 Representing probability of transition of electric automobile from state 4 to state 1, P 42 Representing probability of electric automobile transition from state 4 to state 2, P 43 The probability of the electric automobile transitioning from state 4 to state 3 is represented;
s2.3: calculating a time slot state probability distribution function of the electric automobile according to a Markov transition probability matrix of the electric automobile, wherein the electric automobile state probability distribution function of each time slot in one day is specifically as follows:
wherein ,representing the probability of normal charging state of the electric automobile at the moment t, < >>Probability of representing fast charge state of electric vehicle at time t,/->Representing the probability of the running state of the electric vehicle at the time t, < + > >General for indicating parking state of electric automobile at t momentA rate;
s2.4: obtaining a daily charging demand simulation curve of the electric automobile according to the time slot state probability distribution function of the electric automobile, and obtaining the daily charging demand simulation curve of the electric automobile according to the following formula:
wherein V represents a daily charging demand analog value of the electric automobile, and P n Indicating the power required by the normal charge state of the electric automobile, P f The power required by the fast charge state of the electric automobile is represented, and n represents the number of the electric automobiles in the sampled data;
s3: encoding the daily charging demand simulation curve by using a preset global randomness information encoder, encoding the historical load curve by using a preset local deterministic information encoder, and setting a position code;
the position code is used for marking the sequence of the input transducer encoder in the time dimension, and the calculation formula is as follows:
wherein ,PE(pos,2j) Position coding representing even time dimension, PE (pos,2j+1) Position code representing odd time dimension, pos representing the position of the current slot in the day, d model Representing the overall dimension of the position code; 2j represents even time dimension, 2j+1 represents odd time dimension, satisfying 2 j.ltoreq.d model ，2j+1≤d model ；
S4: fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a preset transducer encoder together with preset auxiliary information to obtain a coding result;
the preset auxiliary information comprises weather information and temperature information;
as shown in fig. 2, the preset transducer encoder includes a first multi-head self-attention layer, a first normalization layer, a first feedforward neural network layer and a second normalization layer which are sequentially connected;
the input end of the first multi-head self-attention layer and the input end of the first normalization layer form residual connection, and the input end of the first feedforward neural network layer and the input end of the second normalization layer form residual connection;
fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a first multi-head self-attention layer together with preset auxiliary information, wherein the input of the first multi-head self-attention layer is marked as a Z vector;
as shown in fig. 3, the Z vector is mapped to Q, K and V matrices by h different linear transforms, respectively, and h attention matrices Z are obtained by self-attention calculation and linear transforms, respectively 1 ；
All attention matrix Z 1 Splicing to obtain a spliced matrix, and combining the spliced matrix with a preset weight matrix W 0 Multiplying to obtain a splice matrix Z containing attention information 2 ；
Splice matrix Z to contain attention information 2 Calculating and outputting through the first normalization layer, the first feedforward neural network layer and the second normalization layer to obtain a coding result, and marking the coding result as Z 3 Vector;
s5: inputting the daily charge demand simulation curve and the encoding result into a preset first converter decoder for decoding to obtain a global information decoding result, and inputting the historical load curve and the encoding result into a preset second converter decoder for decoding to obtain a local information decoding result;
as shown in fig. 4, the first and second transducers decoders are arranged in parallel;
the first transducer decoder comprises a second multi-head self-attention layer, a third normalization layer, a first encoder-decoder attention layer, a fourth normalization layer, a second feedforward neural network layer, a fifth normalization layer and a first linear transformation layer which are sequentially connected;
the input end of the second multi-head self-attention layer is in residual connection with the input end of the third normalization layer, the input end of the first encoder-decoder attention layer is in residual connection with the input end of the fourth normalization layer, and the input end of the second feedforward neural network layer is in residual connection with the input end of the fifth normalization layer;
The second transducer decoder comprises a third multi-head self-attention layer, a sixth normalization layer, a second encoder-decoder attention layer, a seventh normalization layer, a third feedforward neural network layer, an eighth normalization layer and a second linear transformation layer which are sequentially connected;
the input end of the third multi-head self-attention layer is in residual connection with the input end of the sixth normalization layer, the input end of the second encoder-decoder attention layer is in residual connection with the input end of the seventh normalization layer, and the input end of the third feedforward neural network layer is in residual connection with the input end of the eighth normalization layer;
taking a daily charge demand simulation curve as input data of a second multi-head self-attention layer, taking a coding result of a transducer coder and output data of a third normalization layer as input data of a first coder-decoder attention layer, performing linear transformation calculation on output data of a fifth normalization layer through a first linear transformation layer to obtain a global information decoding result, and marking the global information decoding result as Z 4 Vector;
taking the historical load curve as input data of a third multi-head self-attention layer, taking the coding result of a transducer coder and the output data of a sixth normalization layer as input data of a second coder-decoder attention layer, carrying out linear transformation calculation on the output data of an eighth normalization layer through a second linear transformation layer to obtain a local information decoding result, and marking the local information decoding result as Z 5 Vector;
s6: fusing the global information decoding result and the local information decoding result to obtain a daily charging demand prediction curve of the electric automobile;
the first multi-head self-attention layer, the second multi-head self-attention layer and the third multi-head self-attention layer have the same structure;
the first feedforward neural network layer, the second feedforward neural network layer and the third feedforward neural network layer have the same structure and all comprise two full-connection layers.
In a specific implementation process, firstly, electric vehicle initial data of each time slot in a day are acquired, wherein the electric vehicle initial data comprise: state data of the electric automobile, state of charge (SOC) data of a battery of the electric automobile, initial charging time data of the electric automobile and daily driving mileage data of the electric automobile;
the states of the electric automobile include: state 1: normal state of charge, state 2: state of fast charge, state 3: running state and state 4: a parking state;
acquiring a historical load curve of the electric automobile in one day according to the acquired initial data of the electric automobile;
setting basic parameters of an electric automobile, wherein in the embodiment, the energy consumption of a battery of the electric automobile per kilometer is set to be 0.159kWh/km, the power required by a normal charging state is 3.3kW (22V/15A), and the power required by a quick charging state is 50kW (400V/125A);
The maximum capacity of the batteries of different electric vehicles obeys N (mu, sigma) 2 ) Specifically:
wherein f (x) is the maximum capacity of the battery of the xth electric vehicle, where μ=28.5, σ=14.7;
the minimum value of the maximum capacity of the battery of the electric automobile is 10.0, and the maximum value of the maximum capacity of the battery of the electric automobile is 72.0;
in the normal charge state, the quick charge state and the running state of the electric automobile, the state of charge (SOC) of the battery of the electric automobile meets the constraint: SOC is more than or equal to 0.2 and less than or equal to 0.8, so that the service life of a battery of the electric automobile is protected;
randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation, and obtaining a daily charging demand simulation curve of the electric automobile;
the state of each time slot electric automobile in a day is obtained through the scene represented by different SOC values under the daily condition, the total number of the occurrence of different states of each electric automobile in each time slot is added according to the definition of a Markov transition probability matrix, and the occurrence times of different states of the next electric automobile in each time slot are calculated, so that the Markov transition probability matrix of the electric automobile in a day can be statistically deduced for the different states of each electric automobile in each time slot, and the Markov transition probability matrix of the electric automobile is specifically:
Wherein P represents a Markov transition probability matrix of the electric automobile, and P 11 Representing probability of maintaining state 1 of electric automobile in next time slot, P 12 Representing probability of electric automobile transition from state 1 to state 2, P 13 Representing probability of electric automobile transition from state 1 to state 3, P 14 The probability of the electric automobile transitioning from state 1 to state 4 is represented; p (P) 22 Representing probability of maintaining state 2 of electric automobile in next time slot, P 21 Representing probability of transition of electric automobile from state 2 to state 1, P 23 Representing probability of electric automobile transition from state 2 to state 3, P 24 The probability of the electric automobile transitioning from state 2 to state 4 is represented; p (P) 33 Representing probability of maintaining state 3 of electric automobile in next time slot, P 31 Representing probability of transition of electric automobile from state 3 to state 1, P 32 Representing probability of transition of electric automobile from state 3 to state 2, P 44 The probability of the electric automobile transitioning from state 3 to state 4 is represented; p (P) 44 Representing probability of electric automobile maintenance state 4 in next time slot, P 41 Indicating the slave state 4 of the electric automobileProbability of transition to state 1, P 42 Representing probability of electric automobile transition from state 4 to state 2, P 43 The probability of the electric automobile transitioning from state 4 to state 3 is represented;
calculating a time slot state probability distribution function of the electric automobile according to a Markov transition probability matrix of the electric automobile, wherein the electric automobile state probability distribution function of each time slot in one day is specifically as follows:
wherein ,representing the probability of normal charging state of the electric automobile at the moment t, < >>Probability of representing fast charge state of electric vehicle at time t,/->Representing the probability of the running state of the electric vehicle at the time t, < + >>The probability of the parking state of the electric automobile at the moment t is represented;
obtaining a daily charging demand simulation curve of the electric automobile according to the time slot state probability distribution function of the electric automobile, and obtaining the daily charging demand simulation curve of the electric automobile according to the following formula:
wherein V represents a daily charging demand analog value of the electric automobile, and P n Indicating the power required by the normal charge state of the electric automobile, P f The power required by the fast charge state of the electric automobile is represented, and n represents the number of the electric automobiles in the sampled data;
because there is a strong potential correlation between the global randomness information and the local deterministic information, the method in this embodiment uses two encoders to obtain the potential correlation information, specifically:
encoding the daily charge demand simulation curve by using a preset global randomness information encoder, and encoding the historical load curve by using a preset local deterministic information encoder;
because the self-attention mechanism in the transducer neural network cannot confirm the sequence of the input sequence, the embodiment also sets a position code, where the position code is used to mark the sequence of the input transducer encoder in the time dimension, and the calculation formula is as follows:
wherein ,PE(pos,2j) Position coding representing even time dimension, PE (pos,2j+1) Position code representing odd time dimension, pos representing the position of the current slot in the day, d model Representing the overall dimension of the position code; 2j represents even time dimension, 2j+1 represents odd time dimension, satisfying 2 j.ltoreq.d model ，2j+1≤d model ；
Fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a preset transducer encoder together with preset auxiliary information to obtain a coding result;
in this embodiment, the preset auxiliary information includes weather information and temperature information;
fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a first multi-head self-attention layer together with preset auxiliary information, wherein the input of the first multi-head self-attention layer is marked as a Z vector;
mapping the Z vector to Q, K and V matrixes through h different linear transforms respectively, and obtaining h attention matrixes Z through self-attention calculation and linear transforms respectively 1 ；
All attention matrix Z 1 Splicing to obtain a spliced matrix, and combining the spliced matrix with a preset weight matrix W 0 Multiplying to obtain a splice matrix Z containing attention information 2 ；
Splice matrix Z to contain attention information 2 Calculating and outputting through the first normalization layer, the first feedforward neural network layer and the second normalization layer to obtain a coding result, and marking the coding result as Z 3 Vector;
then inputting the daily charge demand simulation curve and the encoding result into a preset first converter decoder for decoding to obtain a global information decoding result, and inputting the historical load curve and the encoding result into a preset second converter decoder for decoding to obtain a local information decoding result, wherein the method specifically comprises the following steps of:
taking a daily charge demand simulation curve as input data of a second multi-head self-attention layer, taking a coding result of a transducer coder and output data of a third normalization layer as input data of a first coder-decoder attention layer, performing linear transformation calculation on output data of a fifth normalization layer through a first linear transformation layer to obtain a global information decoding result, and marking the global information decoding result as Z 4 Vector;
taking the historical load curve as input data of a third multi-head self-attention layer, taking the coding result of a transducer coder and the output data of a sixth normalization layer as input data of a second coder-decoder attention layer, carrying out linear transformation calculation on the output data of an eighth normalization layer through a second linear transformation layer to obtain a local information decoding result, and marking the local information decoding result as Z 5 Vector;
finally decoding the global information to obtain a result Z 4 Vector and local information decoding result Z 5 Vector fusion is carried out to obtain a daily charging power prediction curve of the electric automobile, and the daily charging of the electric automobile is obtainedThe electric demand prediction curve is used for planning the optimal distribution of the electric power system and the economic dispatch of the power grid;
the invention fully focuses on the potential correlation between the global randomness information and the local deterministic information of the daily charging power of the electric automobile, considers the inherent difference between different information, adopts a double-decoder structure, can ensure that decoders of specific tasks keep enough nonlinear fitting capacity and focuses on the prediction characteristics of the respective tasks, so that the finally obtained daily charging demand prediction curve of the electric automobile has the characteristics of high prediction precision and strong robustness, and can obtain stronger prediction performance.
The same or similar reference numerals correspond to the same or similar components;
the terms describing the positional relationship in the drawings are merely illustrative, and are not to be construed as limiting the present patent;
it is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (6)
1. The electric vehicle daily charging demand curve prediction method based on the attention mechanism is characterized by comprising the following steps of:
s1: acquiring initial data of the electric automobile in each time slot in a day, and acquiring a historical load curve of the electric automobile in a day according to the acquired initial data of the electric automobile;
s2: setting basic parameters of an electric automobile, randomly sampling the obtained initial data of the electric automobile by adopting a Monte Carlo algorithm to obtain sampling data of a plurality of time slots of the electric automobile, inputting the sampling data of the plurality of time slots of the electric automobile into a preset Markov model for calculation, and obtaining a daily charging demand simulation curve of the electric automobile, wherein the specific method comprises the following steps:
the basic parameters of the electric automobile comprise: the number, type, maximum battery capacity and state of charge (SOC) constraints of the batteries of the electric vehicle;
s2.1: randomly sampling the obtained initial data of the electric automobile to obtain sampling data of a plurality of time slots of the electric automobile;
s2.2: inputting sampling data of a plurality of time slots of the electric automobile into a preset Markov model, and calculating a Markov transition probability matrix of the electric automobile;
s2.3: calculating a time slot state probability distribution function of the electric automobile according to the Markov transition probability matrix of the electric automobile;
S2.4: acquiring a daily charging demand simulation curve of the electric automobile according to the time slot state probability distribution function of the electric automobile;
s3: encoding the daily charging demand simulation curve by using a preset global randomness information encoder, encoding the historical load curve by using a preset local deterministic information encoder, and setting a position code;
s4: the encoded daily charging demand simulation curve, the encoded historical load curve and the encoded position code are fused, and are input into a preset transducer encoder together with preset auxiliary information, so that an encoding result is obtained, and the specific method comprises the following steps:
the preset auxiliary information comprises weather information and temperature information;
the preset transducer encoder comprises a first multi-head self-attention layer, a first normalization layer, a first feedforward neural network layer and a second normalization layer which are sequentially connected;
the input end of the first multi-head self-attention layer and the input end of the first normalization layer form residual connection, and the input end of the first feedforward neural network layer and the input end of the second normalization layer form residual connection;
fusing the encoded daily charging demand simulation curve, the historical load curve and the position code, and inputting the fused daily charging demand simulation curve, the historical load curve and the position code into a first multi-head self-attention layer together with preset auxiliary information, wherein the input of the first multi-head self-attention layer is marked as a Z vector;
Mapping the Z vector to Q, K and V matrixes through h different linear transforms respectively, and obtaining h attention matrixes Z through self-attention calculation and linear transforms respectively 1 ；
All attention matrix Z 1 Splicing to obtain a spliced matrix, and combining the spliced matrix with a preset weight matrix W 0 Multiplying to obtain a splice matrix Z containing attention information 2 ；
Splice matrix Z to contain attention information 2 Calculating and outputting through the first normalization layer, the first feedforward neural network layer and the second normalization layer to obtain a coding result, and marking the coding result as Z 3 Vector;
s5: the method comprises the steps of inputting a daily charge demand simulation curve and a coding result into a preset first converter decoder to decode to obtain a global information decoding result, and inputting a historical load curve and a coding result into a preset second converter decoder to decode to obtain a local information decoding result, wherein the specific method comprises the following steps of:
the first and second transducers are arranged in parallel;
the first transducer decoder comprises a second multi-head self-attention layer, a third normalization layer, a first encoder-decoder attention layer, a fourth normalization layer, a second feedforward neural network layer, a fifth normalization layer and a first linear transformation layer which are sequentially connected;
The input end of the second multi-head self-attention layer is in residual connection with the input end of the third normalization layer, the input end of the first encoder-decoder attention layer is in residual connection with the input end of the fourth normalization layer, and the input end of the second feedforward neural network layer is in residual connection with the input end of the fifth normalization layer;
the second transducer decoder comprises a third multi-head self-attention layer, a sixth normalization layer, a second encoder-decoder attention layer, a seventh normalization layer, a third feedforward neural network layer, an eighth normalization layer and a second linear transformation layer which are sequentially connected;
the input end of the third multi-head self-attention layer is in residual connection with the input end of the sixth normalization layer, the input end of the second encoder-decoder attention layer is in residual connection with the input end of the seventh normalization layer, and the input end of the third feedforward neural network layer is in residual connection with the input end of the eighth normalization layer;
taking a daily charge demand simulation curve as input data of a second multi-head self-attention layer, taking a coding result of a transducer coder and output data of a third normalization layer as input data of a first coder-decoder attention layer, performing linear transformation calculation on output data of a fifth normalization layer through a first linear transformation layer to obtain a global information decoding result, and marking the global information decoding result as Z 4 Vector;
taking the historical load curve as input data of a third multi-head self-attention layer, taking the coding result of a transducer coder and the output data of a sixth normalization layer as input data of a second coder-decoder attention layer, carrying out linear transformation calculation on the output data of an eighth normalization layer through a second linear transformation layer to obtain a local information decoding result, and marking the local information decoding result as Z 5 Vector;
the first multi-head self-attention layer, the second multi-head self-attention layer and the third multi-head self-attention layer have the same structure;
the first feedforward neural network layer, the second feedforward neural network layer and the third feedforward neural network layer have the same structure and all comprise two full-connection layers;
s6: and fusing the global information decoding result and the local information decoding result to obtain a daily charging demand prediction curve of the electric automobile.
2. The method for predicting the daily charging demand curve of the electric vehicle based on the attention mechanism according to claim 1, wherein the initial data of the electric vehicle in step S1 specifically comprises:
the initial data of the electric automobile comprises: state data of the electric automobile, state of charge (SOC) data of a battery of the electric automobile, initial charging time data of the electric automobile and daily driving mileage data of the electric automobile;
The states of the electric automobile include: state 1: normal state of charge, state 2: state of fast charge, state 3: running state and state 4: and (5) a parking state.
3. The method for predicting the daily charging demand curve of the electric vehicle based on the attention mechanism according to claim 2, wherein in the step S2.2, the markov transition probability matrix of the electric vehicle is specifically:
wherein P represents a Markov transition probability matrix of the electric automobile, and P 11 Representing probability of maintaining state 1 of electric automobile in next time slot, P 12 Representing probability of electric automobile transition from state 1 to state 2, P 13 Representing probability of electric automobile transition from state 1 to state 3, P 14 The probability of the electric automobile transitioning from state 1 to state 4 is represented; p (P) 22 Representing probability of maintaining state 2 of electric automobile in next time slot, P 21 Representing probability of transition of electric automobile from state 2 to state 1, P 23 Representing probability of electric automobile transition from state 2 to state 3, P 24 The probability of the electric automobile transitioning from state 2 to state 4 is represented; p (P) 33 Representing probability of maintaining state 3 of electric automobile in next time slot, P 31 Representing probability of transition of electric automobile from state 3 to state 1, P 32 Representing probability of transition of electric automobile from state 3 to state 2, P 34 The probability of the electric automobile transitioning from state 3 to state 4 is represented; p (P) 44 Representing probability of electric automobile maintenance state 4 in next time slot, P 41 Representing probability of transition of electric automobile from state 4 to state 1, P 42 Representing probability of electric automobile transition from state 4 to state 2, P 43 The probability of the electric vehicle transitioning from state 4 to state 3 is shown.
4. The method for predicting a daily charging demand curve of an electric vehicle based on an attention mechanism according to claim 3, wherein in step S2.3, a time slot state probability distribution function of the electric vehicle is specifically:
the electric automobile state probability distribution function of each time slot in one day is specifically:
wherein ,representing the probability of normal charging state of the electric automobile at the moment t, < >>Probability of representing fast charge state of electric vehicle at time t,/->Representing the probability of the running state of the electric vehicle at the time t, < + >>And the probability of the parking state of the electric automobile at the time t is shown.
5. The method for predicting the daily charging demand curve of the electric vehicle based on the attention mechanism according to claim 4, wherein in the step S2.4, the specific method for obtaining the daily charging demand simulation curve of the electric vehicle according to the time slot state probability distribution function of the electric vehicle is as follows:
Obtaining a daily charging demand simulation curve of the electric automobile according to the following formula:
wherein V represents the daily charging requirement of the electric automobileAnalog value, P n Indicating the power required by the normal charge state of the electric automobile, P f The power required by the electric automobile in the quick charge state is represented, and n represents the number of the electric automobiles in the sampled data.
6. The method for predicting the daily charging demand curve of the electric automobile based on the attention mechanism according to claim 1, wherein the position code in the step S3 is specifically:
the position code is used for marking the sequence of the input transducer encoder in the time dimension, and the calculation formula is as follows:
wherein ,PE(pos,2j) Position coding representing even time dimension, PE (pos,2j+1) Position code representing odd time dimension, pos representing the position of the current slot in the day, d model Representing the overall dimension of the position code; 2j represents even time dimension, 2j+1 represents odd time dimension, satisfying 2 j.ltoreq.d model ，2j+1≤d model 。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211415516.3A CN115730710B (en) | 2022-11-11 | 2022-11-11 | Electric vehicle daily charging demand curve prediction method based on attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211415516.3A CN115730710B (en) | 2022-11-11 | 2022-11-11 | Electric vehicle daily charging demand curve prediction method based on attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115730710A CN115730710A (en) | 2023-03-03 |
| CN115730710B true CN115730710B (en) | 2023-09-08 |
Family
ID=85295404
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211415516.3A Active CN115730710B (en) | 2022-11-11 | 2022-11-11 | Electric vehicle daily charging demand curve prediction method based on attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115730710B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295860A (en) * | 2016-07-29 | 2017-01-04 | 国网山东省电力公司经济技术研究院 | A kind of electric automobile scale charge requirement Forecasting Methodology based on Monte Carlo Analogue Method |
| KR20190035637A (en) * | 2019-03-22 | 2019-04-03 | 주식회사 알고리고 | Apparatus for estimating power demand using attention matrix |
| KR20210105793A (en) * | 2020-02-19 | 2021-08-27 | (주)뤼이드 | A system for predicting user drop out rate based on artificial intelligence learning and method thereof |
| CN113592185A (en) * | 2021-08-05 | 2021-11-02 | 四川大学 | Power load prediction method based on Transformer |
| CN114331542A (en) * | 2021-12-30 | 2022-04-12 | 智光研究院(广州)有限公司 | Method and device for predicting charging demand of electric vehicle |
| CN114548517A (en) * | 2022-01-21 | 2022-05-27 | 广州蔚景科技有限公司 | Estimation and modeling method for charging load of electric automobile |
-
2022
- 2022-11-11 CN CN202211415516.3A patent/CN115730710B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106295860A (en) * | 2016-07-29 | 2017-01-04 | 国网山东省电力公司经济技术研究院 | A kind of electric automobile scale charge requirement Forecasting Methodology based on Monte Carlo Analogue Method |
| KR20190035637A (en) * | 2019-03-22 | 2019-04-03 | 주식회사 알고리고 | Apparatus for estimating power demand using attention matrix |
| KR20210105793A (en) * | 2020-02-19 | 2021-08-27 | (주)뤼이드 | A system for predicting user drop out rate based on artificial intelligence learning and method thereof |
| CN113592185A (en) * | 2021-08-05 | 2021-11-02 | 四川大学 | Power load prediction method based on Transformer |
| CN114331542A (en) * | 2021-12-30 | 2022-04-12 | 智光研究院(广州)有限公司 | Method and device for predicting charging demand of electric vehicle |
| CN114548517A (en) * | 2022-01-21 | 2022-05-27 | 广州蔚景科技有限公司 | Estimation and modeling method for charging load of electric automobile |
Non-Patent Citations (1)
| Title |
|---|
| Self-Attention-Based Machine Theory of Mind for Electric Vehicle Charging Demand Forecast;Tianyu Hu 等;《IEEE Transactions on Industrial Informatics》;第18卷(第11期);8191-8202 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115730710A (en) | 2023-03-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Deep-learning-based probabilistic forecasting of electric vehicle charging load with a novel queuing model | |
| CN102779223B (en) | The method of short-term electric load prediction and device | |
| Pertl et al. | An equivalent time-variant storage model to harness EV flexibility: Forecast and aggregation | |
| WO2022141213A1 (en) | Gene prediction method and system for fault of autonomous rail rapid transit vehicle in smart city | |
| Cheng et al. | Electric bus fast charging station resource planning considering load aggregation and renewable integration | |
| CN106203693A (en) | A kind of system and method for Power Output for Wind Power Field climbing event prediction | |
| CN111191856A (en) | Regional comprehensive energy system multi-energy load prediction method considering time sequence dynamic characteristics and coupling characteristics | |
| CN116187548A (en) | Photovoltaic power generation power prediction method and device, storage medium and electronic device | |
| CN113627661A (en) | Method for predicting charging load of electric automobile | |
| CN116151114A (en) | Method and system for predicting service life of fuel cell under meta-universe based on hybrid framework | |
| Wang et al. | Short‐term passenger flow forecasting using CEEMDAN meshed CNN‐LSTM‐attention model under wireless sensor network | |
| CN113988373B (en) | Multi-task massive user load prediction method based on multi-channel convolutional neural network | |
| CN116433223A (en) | Substation equipment fault early warning method and equipment based on double-domain sparse transducer model | |
| Yan et al. | Data‐driven robust planning of electric vehicle charging infrastructure for urban residential car parks | |
| CN115730710B (en) | Electric vehicle daily charging demand curve prediction method based on attention mechanism | |
| CN110648013A (en) | Electric vehicle charging load prediction method based on dual-mode maximum entropy | |
| Li et al. | Short-mid term electricity consumption prediction using non-intrusive attention-augmented deep learning model | |
| CN111509782B (en) | Probabilistic power flow analysis method considering charging load and photovoltaic output random characteristics | |
| CN109711612A (en) | A kind of Wind power forecasting method and device optimizing echo state network | |
| Li et al. | DiffCharge: Generating EV Charging Scenarios via a Denoising Diffusion Model | |
| CN116596169A (en) | Power system prediction method, device and storage medium | |
| Liu et al. | Electric Vehicle Load Forecast Based on Higher Order Markov Chain | |
| CN112001518A (en) | Prediction and energy management method and system based on cloud computing | |
| CN117494882B (en) | Urban multi-scene charging load prediction method based on vehicle operation background data | |
| Yin et al. | Research on short-term power load prediction based on deep 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |