CN117559468A - V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction - Google Patents
V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction Download PDFInfo
- Publication number
- CN117559468A CN117559468A CN202311061524.7A CN202311061524A CN117559468A CN 117559468 A CN117559468 A CN 117559468A CN 202311061524 A CN202311061524 A CN 202311061524A CN 117559468 A CN117559468 A CN 117559468A
- Authority
- CN
- China
- Prior art keywords
- ultra
- time
- short
- term
- frequency
- 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
- 230000004044 response Effects 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000004364 calculation method Methods 0.000 claims abstract description 27
- 230000008859 change Effects 0.000 claims abstract description 9
- 238000011217 control strategy Methods 0.000 claims abstract description 9
- 238000005070 sampling Methods 0.000 claims description 32
- 230000015654 memory Effects 0.000 claims description 29
- 238000013528 artificial neural network Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 21
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000000611 regression analysis Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 230000007246 mechanism Effects 0.000 claims description 7
- 230000000694 effects Effects 0.000 claims description 6
- 238000012417 linear regression Methods 0.000 claims description 6
- 230000000737 periodic effect Effects 0.000 claims description 6
- 230000001932 seasonal effect Effects 0.000 claims description 6
- 230000001105 regulatory effect Effects 0.000 claims description 5
- 230000001276 controlling effect Effects 0.000 claims description 4
- 230000006978 adaptation Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000005314 correlation function Methods 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000012886 linear function Methods 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 239000003795 chemical substances by application Substances 0.000 description 18
- 238000004146 energy storage Methods 0.000 description 15
- 238000010586 diagram Methods 0.000 description 14
- 230000008901 benefit Effects 0.000 description 6
- 238000007599 discharging Methods 0.000 description 3
- 238000004880 explosion Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- 238000007665 sagging Methods 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000009194 climbing Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000013101 initial test Methods 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/24—Arrangements for preventing or reducing oscillations of power in networks
- H02J3/241—The oscillation concerning frequency
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/28—Arrangements for balancing of the load in a network by storage of energy
- H02J3/32—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means
- H02J3/322—Arrangements for balancing of the load in a network by storage of energy using batteries with converting means the battery being on-board an electric or hybrid vehicle, e.g. vehicle to grid arrangements [V2G], power aggregation, use of the battery for network load balancing, coordinated or cooperative battery charging
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the technical field of frequency modulation at the load side of a power grid, in particular to a V2G station rapid frequency modulation response method based on ultra-short-term frequency deviation prediction, which comprises the following steps: constructing an ultra-short-term load prediction model based on historical load data of a power grid; predicting future load change conditions every 5 minutes in a rapid frequency modulation response period through an ultra-short-term load prediction model; analyzing the relationship between the ultra-short-term load fluctuation rate and the system frequency based on the historical load data and the frequency statistical data of the power grid; calculating the ultra-short-term load fluctuation rate according to the ultra-short-term prediction result, and performing ultra-short-term frequency deviation prediction calculation; acquiring information of each charging pile of the current V2G station; carrying out power distribution adjustment on each charging pile based on ultra-short-term frequency deviation prediction; and acquiring the power grid frequency value of the charging and replacing power station in real time, calculating the real-time frequency deviation and transmitting the real-time frequency deviation to each charging pile. According to the invention, the advanced power distribution adjustment according to the charging pile and participation in the primary frequency modulation response control strategy of the power grid are realized, and the response time and the frequency response speed are improved.
Description
Technical Field
The invention relates to the technical field of frequency modulation at a load side of a power grid, in particular to a V2G station rapid frequency modulation response method based on ultra-short-term frequency deviation prediction.
Background
The weak inertia, intermittence and randomness of the new energy source can seriously affect the electric energy quality of the power grid and the power supply reliability of the power system, so that the problem of insufficient frequency modulation capacity of the system is increasingly outstanding, and higher requirements are put forward for frequency modulation of the power grid.
The traditional frequency modulation mode is carried out around the generator set, the output of the generator set is regulated to balance the load demand, the conventional set participates in the frequency modulation of the power grid, the problems of slow response, low climbing speed and the like exist, the defects can be overcome by the rapid response and accurate control characteristics of the V2G station, the V2G station is more complicated in participation in a rapid frequency modulation scene compared with a conventional energy storage battery due to different time charging demands, and the key problem of the frequency modulation application of the V2G station is how to formulate a reasonable power distribution strategy of the participation of the V2G station in the rapid frequency modulation under different scenes.
Based on the above situation, the present disclosure provides a fast frequency modulation response method for a V2G station based on ultra-short term frequency deviation prediction, based on the ultra-short term load prediction result, estimates the power grid frequency deviation and primary frequency modulation requirement in the next stage, and adjusts the power distribution in the current state of the current V2G station in advance, so that the V2G station corrects the system frequency fluctuation in the next stage better.
Disclosure of Invention
Based on the above purpose, the invention provides a V2G station fast frequency modulation response method based on ultra-short term frequency deviation prediction.
A V2G station fast frequency modulation response method based on ultra-short term frequency deviation prediction comprises the following steps:
step one: constructing an ultra-short-term load prediction model based on historical load data of a power grid;
step two: predicting load change conditions every 5 minutes in a rapid frequency modulation response period through an ultra-short-term load prediction model, wherein the prediction time is 5 minutes to 1 hour in the future;
step three: analyzing the relationship between the ultra-short-term load fluctuation rate and the system frequency based on the historical load data and the frequency statistical data of the power grid;
step four: calculating the ultra-short-term load fluctuation rate according to the ultra-short-term prediction result, and performing ultra-short-term frequency deviation prediction calculation;
step five: acquiring charge and discharge power, a vehicle SOC state, a start SOC, a stop SOC, a user demand mileage, a maximum power constraint of battery charge and discharge, a user lifting time and user participation response willingness information of each charging pile of a current V2G station;
step six: carrying out power distribution adjustment on each charging pile based on ultra-short-term frequency deviation prediction, reserving a phase of response power grid frequency modulation demand capacity according to a charging pile calculation result, and updating a charging pile frequency modulation virtual sagging control strategy;
step seven: and acquiring the power grid frequency value of the charging and replacing power station in real time, calculating the real-time frequency deviation, synchronously transmitting the real-time frequency deviation to each charging pile, and regulating the charging power by each charging pile foundation according to the real-time frequency deviation.
Further, the ultra-short term load prediction model is based on a neural network algorithm, a time sequence algorithm and a regression analysis algorithm.
Further, the neural network algorithm is specifically a cyclic neural network, the regression analysis algorithm is specifically a multiple linear regression algorithm, and the time sequence algorithm comprises a two-way long-short-term memory neural network algorithm and a Facebook-Prophet algorithm.
Further, the multiple linear regression algorithm considers the model of time and weather independent variable factors as follows:
Load＝β 0 +β 1t +β 2w +β 3 3+...+β n×n +ε
where β0, β1,..βn is a model parameter, learning from the data is needed, ε is an error term, t is time, and w is weather.
Further, the Facebook-propset algorithm breaks down the time series in its entirety into 3 parts: the effects of growing trend, seasonal trend and holiday, the specific function decomposition is as follows:
y(t)＝g(t)+s(t)+h(t)+∈ t
wherein g (t) represents a growth function for fitting non-periodic variations of the predicted values in the time series; s (t) is a period term for representing the periodic variation of the time series data, h (t) is a holiday term for representing the influence of holiday and holiday reasons on the time series data, and the specific correlation function is described as follows:
the growth function is classified into a linear growth and a nonlinear growth, the linear growth is realized by using a piecewise linear function, and the formula is:
g(t)＝(k+a(t)δ)·t+(m+a(t) T γ)
the nonlinear growth is implemented using a logic function, the formula of which is as follows:
where k represents the growth rate, δ is the adaptation rate, m is the offset parameter, C (t) For maximum bearing capacity, defining maximum value of growth, a (t) is vector, and gamma is offset parameter;
the period term is expressed using a fourier series, the formula of which is as follows:
wherein P represents the period of the time sequence, 2n represents the number of periods used in the model, and when n is too large, a complex seasonal function is fitted;
holiday items treat different holidays as independent models at different time points, and a time window is set up for each holiday model, and the holiday model is expressed as:
k＝(k 1 ，…,k L ) T
wherein i represents holidays, di represents time t contained in a window period, ki represents influence of holidays on a prediction result, and 1 represents current value of Di.
Further, the two-way long-short-term memory neural network algorithm comprises a gating mechanism, each two-way long-short-term memory neural network algorithm calculation unit comprises 3 control gates, namely an input gate, an output gate and a forget gate, and the specific algorithm is as follows:
forgetting door f t For forgetting the upper memory cell state c t-1 Information:
f t ＝σ(W f +U f h t-1 +b f )
wherein: w (W) f Is a weight matrix of forgetting gates; b f Is the forgetting door offset; sigma is a Sigmoid function;
candidate state of memory cellThe calculation of (a) is shown as the following formula, input gate i t Determining information retained in candidate states of the current cell:
i t ＝σ(W i +U i h t-1 +b i )
wherein: w (W) i i、W c Respectively represent input gates i t And candidate statesIs a weight matrix of (2); b i 、b c Is the corresponding offset;
through i t And f t Combining the previous time memory state c t-1 And a current time candidate memory stateTo update the memory cell state c at the current time t ：
Wherein Θ represents multiplication by element;
output door o t The method is mainly used for controlling the output of the state value of the memory cell:
o t ＝σ(W o x t +U o h t-1 +b 0 )
wherein: w (W) o Is an output gate o t Weight matrix of (a); b 0 Is the offset of the output gate;
the hidden layer output value ht is obtained through nonlinear calculation:
h t ＝o t Θtanh(c t )。
in the fourth step, a sliding average filtering method or a wavelet analysis method is adopted to extract the high-frequency load fluctuation rate.
Furthermore, the method also comprises an unequal weight combination algorithm prediction model, wherein the unequal weight combination algorithm prediction model is based on a neural network algorithm, a time sequence algorithm and a regression analysis algorithm, and the unequal weight combination algorithm prediction model is used for giving optimal solution to obtain a final prediction model.
Furthermore, the unequal weight combination algorithm prediction model specifically combines the results of a plurality of single prediction models, the weights of the single prediction models in final prediction are different, and for each single prediction model, the weights of the prediction results are adjusted based on criteria, wherein the criteria comprise the historical prediction performance of the model and the complexity of the model.
Furthermore, the high-frequency load fluctuation rate adopts a sliding average value filtering method, and specifically comprises the following steps: firstly, a data sampling window is established in a RAM, M pieces of sampling data are stored, arithmetic average is carried out on the data in the sampling window, then the sampling window is shifted backwards by a certain length, each time new data is acquired, the earliest acquired data is lost, and average calculation is repeated;
assuming that the time length of the moving average sampling window is Mmin, the high frequency component PHt and the low frequency component PLt of the load at time t are calculated by the following formula:
wherein: PLt is the load low frequency component; PHt is the load high frequency component; pt is the load of tmin measured; t is a sampling time point; n is the total number of sampling points.
The invention has the beneficial effects that:
according to the invention, the advanced power distribution adjustment according to the charging pile and the primary frequency modulation response control strategy of the power grid are realized based on ultra-short-term power prediction by considering the SOC requirements of users and the willingness to participate in frequency modulation response, so that the response time and the frequency response speed are improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it being obvious that the drawings in the description below are only of the invention and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a fast frequency response flow chart according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an input gate, an output gate, and a forget gate in an LSTM computation unit according to an embodiment of the invention;
FIG. 3 is a schematic diagram of a forward and reverse LSTM structure according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a fuzzy empirical function according to an embodiment of the present invention;
FIG. 5 is a schematic diagram showing filtering effects of different sliding durations according to an embodiment of the present invention;
FIG. 6 is a graph of the high frequency components of the load for the sliding duration of an embodiment of the present invention;
fig. 7 is a schematic diagram of a hierarchical management structure of an electric automobile participant according to an embodiment of the present invention;
FIG. 8 is a schematic diagram of a secondary agent control center according to an embodiment of the present invention;
FIG. 9 is a schematic diagram of control when the distributed power for frequency modulation is 60kW according to an embodiment of the present invention;
FIG. 10 is a control schematic diagram of the frequency modulation distributed power of-60 kW according to the embodiment of the invention;
FIG. 11 is a schematic diagram of a total energy storage model according to an embodiment of the present invention;
fig. 12 is a schematic diagram of a participation frequency modulation model considering user requirements of an electric vehicle according to an embodiment of the present invention;
fig. 13 is a schematic diagram of frequency droop control characteristics of an electric vehicle according to an embodiment of the present invention;
fig. 14 is a schematic diagram of power setting control of an electric vehicle participating in frequency modulation according to an embodiment of the present invention.
Detailed Description
The present invention will be further described in detail with reference to specific embodiments in order to make the objects, technical solutions and advantages of the present invention more apparent.
It is to be noted that unless otherwise defined, technical or scientific terms used herein should be taken in a general sense as understood by one of ordinary skill in the art to which the present invention belongs. The terms "first," "second," and the like, as used herein, do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. The word "comprising" or "comprises", and the like, means that elements or items preceding the word are included in the element or item listed after the word and equivalents thereof, but does not exclude other elements or items. The terms "connected" or "connected," and the like, are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "upper", "lower", "left", "right", etc. are used merely to indicate relative positional relationships, which may also be changed when the absolute position of the object to be described is changed.
As shown in fig. 1-14, a fast frequency modulation response method for a V2G station based on ultra-short term frequency deviation prediction includes the following steps:
step one: constructing an ultra-short-term load prediction model based on historical load data of a power grid;
step two: predicting load change conditions every 5 minutes in a rapid frequency modulation response period through an ultra-short-term load prediction model, wherein the prediction time is 5 minutes to 1 hour in the future;
step three: analyzing the relationship between the ultra-short-term load fluctuation rate and the system frequency based on the historical load data and the frequency statistical data of the power grid;
step four: calculating the ultra-short-term load fluctuation rate according to the ultra-short-term prediction result, and performing ultra-short-term frequency deviation prediction calculation;
step five: acquiring charge and discharge power, a vehicle SOC state, a start SOC, a stop SOC, a user demand mileage, a maximum power constraint of battery charge and discharge, a user lifting time and user participation response willingness information of each charging pile of a current V2G station;
step six: carrying out power distribution adjustment on each charging pile based on ultra-short-term frequency deviation prediction, reserving a phase of response power grid frequency modulation demand capacity according to a charging pile calculation result, and updating a charging pile frequency modulation virtual sagging control strategy;
step seven: the method comprises the steps of acquiring a power grid frequency value of a charging and replacing power station in real time, calculating real-time frequency deviation and synchronously transmitting the real-time frequency deviation to each charging pile, adjusting charging power by each charging pile foundation according to the real-time frequency deviation, namely judging a power adjusting direction and adjusting power according to the real-time frequency deviation and a control strategy that the current charging pile participates in primary frequency modulation virtual sagging of the power grid when the power grid frequency exceeds a preset primary frequency modulation dead zone, so that frequency modulation response of partial charging piles considering the SOC requirement of a user is realized, and starting time and frequency response speed of the frequency modulation response are improved.
The ultra-short term load prediction model is based on a neural network algorithm, a time sequence algorithm and a regression analysis algorithm.
The neural network algorithm is specifically a cyclic neural network (RNN), the regression analysis algorithm is specifically a multiple linear regression algorithm, and the time sequence algorithm comprises a two-way long-short-term memory neural network algorithm and a Facebook-Prophet algorithm.
The multiple linear regression algorithm takes into account the time and weather independent factors as follows:
Load＝β0+β1t+β2w+β3x3+...+βnxn+ε
where β0, β1,..βn is a model parameter, learning from the data is needed, ε is an error term, t is time, and w is weather.
The Facebook-propset algorithm breaks down the time series in its entirety into 3 parts: the effects of growing trends, seasonal trends, and holidays. The specific functional decomposition is as follows:
y(t)＝g(t)+s(t)+h(t)+∈ t
wherein g (t) represents a growth function for fitting non-periodic variations of the predicted values in the time series; s (t) is a period term for representing the periodic variation of the time series data, h (t) is a holiday term for representing the influence of holiday and holiday reasons on the time series data, and the specific correlation function is described as follows:
the growth function is divided into linear growth and nonlinear growth, the linear growth is realized by using piecewise linear function, and the formula is:
g(t)＝(k+a(t)δ)·t+(m+a(t) T γ)
the nonlinear growth is implemented using a logic function, the formula of which is as follows:
where k represents the growth rate, δ is the adaptation rate, m is the offset parameter, C (t) For maximum bearing capacity, defining maximum value of growth, a (t) is vector, and gamma is offset parameter;
the period term is expressed using a fourier series, the formula of which is as follows:
where P represents the period of the time series and 2n represents the number of periods used in the model. If n is larger, a complex seasonal function can be fitted, but an overfitting phenomenon can also occur;
holiday items treat different holidays as independent models at different time points, and a time window is set up for each holiday model, and the holiday model is expressed as:
K＝(k 1 ，…,k L ) T
wherein i represents holidays, di represents time t contained in a window period, ki represents influence of holidays on a prediction result, and 1 represents current value of Di.
The two-way long-short-period memory neural network algorithm comprises a gating mechanism, each two-way long-short-period memory neural network algorithm calculation unit comprises 3 control gates, namely an input gate, an output gate and a forgetting gate, and the problems of RNN gradient explosion and gradient disappearance are solved to a certain extent due to the addition of the gating mechanism.
In order to selectively update the memory unit, the LSTM introduces a unit state based on the original RNN hidden layer state for long-term memory, reflecting the dependency relationship of adjacent moments learned by the deep network at any time step and the structural features of the longer-time previously input data. Each LSTM computing unit comprises 3 control gates, namely an input gate, an output gate and a forget gate, the structure of which is shown in figure 2,
let the input sequence be { x } 1 ，x 2 ，…，x T X, where x t ＝{x t，1 ，x t，2 ，…，x t，k }∈R k Representing k-dimensional true vector data at the t-th time step. The LSTM is utilized to construct a load prediction model, and the internal updating process of each unit is as follows:
forgetting door f t For forgetting the upper memory cell state c t-1 Information:
f t ＝σ(W f +U f h t-1 +b f )
wherein: w (W) f Is a weight matrix of forgetting gates; b f Is the forgetting door offset; sigma is typically a Sigmoid function;
candidate state of memory cellThe calculation of (a) is shown as the following formula, input gate i t Determining information retained in candidate states of the current cell:
i t ＝σ(W i +U i h t-1 +b i )
wherein: w (W) i i、W c Respectively represent input gates i t And candidate statesIs a weight matrix of (2); b i 、b c Is the corresponding offset;
through i t And f t Combining the previous time memory state ct- 1 And a current time candidate memory stateTo update the memory cell state c at the current time t ：
Wherein Θ represents multiplication by element;
output door o t The method is mainly used for controlling the output of the state value of the memory cell:
o t ＝σ(W o x t +U o h t-1 +b 0 )
wherein: w (W) o Is an output gate o t Weight matrix of (a); b 0 Is the offset of the output gate;
the hidden layer output value ht is obtained through nonlinear calculation:
ht＝otΘtanh(ct)
the unit weight and bias of each control cell information flow in the above formulas are used for carrying out load prediction on time sequence through training learning, and usually, the information in the LSTM network is transmitted in one way, only past information can be used, but future information cannot be used, and in order to adapt to the peak load amplitude change characteristics of the day, a two-way long-short-term memory neural network (Bi-LSTM) is selected to be used for constructing a prediction model.
The Bi-LSTM is formed by combining a forward LSTM and a reverse LSTM, and the structure is shown in figure 3. The forward LSTM can acquire the past data information of the input sequence, the backward LSTM can acquire the future data information of the input sequence, and the forward and backward LSTM training is realized on the time sequence, so that the global property and the integrity of feature extraction are further improved.
At t time instant, the hidden layer output value H of Bi-LSTM t From the forward directionAnd reverse->The composition is as follows:
the method comprises the steps of initializing a plurality of network parameters including input dimension, output dimension, iteration round number and activation function, constructing a network structure, obtaining daily peak load through a full-connection layer as prediction output of a model, and optimizing the model parameters in an iteration training mode.
The invention also fuses Prophet and Bi-LSTM algorithm to predict the load of iron and steel enterprises;
the Bi-LSTM algorithm has strong time sequence, can fully mine the rule of historical data, and has the defects of gradient elimination and explosion; the Prophet algorithm has the advantages of small-scale sample optimization and linear mapping, the problems of gradient disappearance and explosion can be solved to a certain extent, the weight distribution model adopts a fuzzy set logic algorithm, the difference ratio of the predicted result of the Bi-LSTM and Prophet algorithm to the predicted value at the previous moment is taken, and after mathematical fitting of multiple calculation experiments, the error of the predicted result reaches the minimum value when the fuzzy empirical function takes the value as shown in figure 4.
Assume that the Bi-LSTM and Prophet predictions are W, respectively Bi-LSTM And W is Prophet Its final prediction result
The weight of each type of prediction model is a 1i And a 2i . Let T 1i ＝|W iBi-LSTM -W i-1 |，T 2i ＝|W iProphet -W i-1 I, wherein W i-1 A is the load predicted value of the last moment 1i And a 2i Is determined by the following formula;
a 2i ＝1-a 1i
and step four, extracting the high-frequency load fluctuation rate by adopting a sliding average value filtering method or a wavelet analysis method.
The method also comprises an unequal weight combination algorithm prediction model, wherein the unequal weight combination algorithm prediction model is based on a neural network algorithm, a time sequence algorithm and a regression analysis algorithm, and the unequal weight combination algorithm prediction model is endowed with unequal weight solution optimization to obtain a final prediction model.
The unequal weight combination algorithm prediction model specifically combines the results of a plurality of single prediction models, the weights of the plurality of single prediction models in final prediction are different, and for each single prediction model, the weights of the prediction results are adjusted based on criteria, wherein the criteria comprise historical prediction performance of the model and complexity of the model.
The high-frequency load fluctuation rate adopts a sliding average value filtering method, and specifically comprises the following steps: firstly, a data sampling window is established in a RAM, M pieces of sampling data are stored, arithmetic average is carried out on the data in the sampling window, then the sampling window is shifted backwards by a certain length, each time new data is acquired, the earliest acquired data is lost, and average calculation is repeated;
assuming that the time length of the moving average sampling window is M min, (M is an even number), the high-frequency component PHt and the low-frequency component PLt of the load at the time t are calculated by the following formula:
wherein: PLt is the load low frequency component; PHt is the load high frequency component; pt is the load of the measured t min; t is a sampling time point; n is the total number of sampling points.
The length of the sampling window time of the moving average can be arbitrarily selected, the filtering effect of the sampling window time is related to the sampling window time,
if the sampling time window is selected too long, the variation trend of the load will react on the load high-frequency component PHt, so that the load high-frequency component PHt is no longer random, while if the sampling time window is too short, the randomness of the load variation will react on the load low-frequency component PLt, so that selecting a suitable sampling time window is the key of the sliding average method, taking a certain day of load data in a certain area as an example (rated load of 100 MW), and the sliding average sampling time is respectively selected from 5min, 15min, 30min and 60min, as shown in fig. 5.
From the above graph, it can be seen that the longer the sliding average sampling time is, the smoother the obtained load low-frequency component is, but when the sliding sampling time is taken for 60min, the change trend of the obtained load low-frequency component is obviously deviated from the original load change trend, and when the sliding sampling time is taken for 15min, the obtained load low-frequency component still has stronger random volatility, and 30min is selected as the sliding average sampling time, and the obtained load high-frequency component is shown in fig. 6.
The following are specific embodiments of the present invention:
the invention assumes that an electric automobile agent exists in the future, and can uniformly manage the electric automobile charging system adopting the uniform charging pile standard and the same communication standard, wherein the management adopts a layered structure, as shown in figure 7
Corresponding to each frequency adjustment control zone, there is a zone level one agent whose role is to distribute the power signal. The frequency modulation capacity is calculated according to the number of controllable electric vehicles of each next-level agent, the SOC of each electric vehicle and the running requirement of a vehicle owner. In addition, the centralized charging station/power exchange station is also suitable for an agent mechanism, and the centralized charging station/power exchange station of the electric vehicle is considered to be managed by a secondary agent when participating in frequency modulation, and the centralized charging station/power exchange station of the electric vehicle participates in the frequency modulation process of the electric power system and can be considered to be a centralized energy storage device. At present, the benefits of the traditional unit participating in primary frequency modulation are reflected in multiple electric energy settlement, and the benefits obtained by secondary frequency modulation are reflected in spare capacity benefits and multiple electric energy benefits responding to an AGC signal. For the electric automobile to participate in frequency modulation, the electric energy generated by participating in primary frequency modulation needs to be settled according to a protocol signed by a dispatching mechanism of an agent and a frequency modulation control area, and the settlement price of the electric energy needs to consider the two aspects of power response and battery life loss, which are provided by the electric automobile participating in frequency modulation and are superior to those of the traditional unit, so that the electric automobile has more excitation for electric automobile users, and more electric automobile users can participate in the frequency modulation service of an electric system; the same consideration is also given to secondary frequency modulation, and even charging electricity prices are policy-oriented toward users participating in frequency modulation. Similar to the control manner of the traditional power generation unit, the power electronic interface of the single electric automobile can react to the frequency deviation to participate in primary frequency modulation. And during secondary frequency modulation, the AGC sends a responsive frequency modulation power response signal to the traditional generator and the electric automobile agent. The electric automobile agent distributes power adjusted in response to the frequency according to the number of electric automobiles in a certain range.
The cars traveling on the road are much smaller than the stopped cars, which means that substantially most of the cars are stopped. This situation will not change with the development of electric vehicles in the future. The unit energy storage of the electric automobile is less than that of a gasoline automobile, and the electric automobile is often required to be charged. But the electric automobile is convenient to charge, and can be charged by using lower charging power at the place where the alternating current power supply is provided. Because of concerns of vehicle owners about driving mileage, the electric vehicle is charged frequently, which results in that the SOC of the battery of the electric vehicle is basically close to 100% during charging, the power distribution strategy studied herein is that the electric vehicle is charged to 85%, the SOC of the electric vehicle is involved in the frequency modulation control between 80% and 90%, according to the literature, the SOC of the electric vehicle is regulated close to 100% to shorten the battery life faster, and the SOC of the vehicle should be above 80% when the vehicle is disconnected from the power grid, considering the demands of users.
The secondary agent distributes the frequency modulated power signal to the electric vehicle while receiving information of the electric vehicle. As shown in fig. 8, there are 500 secondary agent control centers, each agent controlling 100 electric vehicles in its jurisdiction, the number of electric vehicles controlled being 50000.
Each secondary agent control center transmits a power distribution signal into the electric vehicle every second. The charging and discharging power of the controllable electric automobile is limited according to the charging power and the controllable capacity. Each electric car uploads whether, when, and when it is controllable to the secondary agent control center.
Controllable electric automobile energy storage calculation:
under the agency mechanism, the available energy storage of the electric automobile at any time in the day can be represented by a formula (4-1):
E control (t)＝E 0 +E control-in (t)-E plug-out (t)-E LFC (t) (4-1)
wherein E is 0 Is the initial energy storage of the controllable electric automobile, E plug-in (t) the energy storage increment caused by the electric automobile changing into a controllable state, and the calculation formulas of the energy storage increment are as follows;
E plug-out and (t) is the energy storage reduction amount caused when the electric automobile is disconnected from the power grid. The SOC control strategy is adopted herein, so that the battery SOC of the electric vehicle in the controlled state can be approximately considered to be equal to the average SOC of all the controlled electric vehicles. Thus E is plug-out The calculation of (t) may use a common algorithmCalculation of formula (4-4):
in E LFC (t) is t 0 And (3) in the period of t, the controllable electric automobile participates in energy storage variable quantity caused by frequency modulation of the system, and the calculation formula is as follows:
p in the formula LFC And (t) is the resulting frequency modulated power allocation by the secondary agent, the sign of which is specified as positive discharge. The calculation formulas of the SOC of all the controllable electric automobiles are as follows:
E sum and (t) represents the total capacity of the controllable electric automobile in the area controlled by the secondary agent, and the calculation formula is as follows:
in summary, a total energy storage model is shown in fig. 11.
Power and energy constraints:
the battery SOC of the controllable electric automobile participating in the frequency modulation of the electric power system is [80%,90%]Range fluctuation, the total energy storage of which is also in (E cmin (t)，E cmax (t)) range. The calculation formulas of the upper limit and the lower limit of the total energy storage are as follows:
in SOC min And SOC (System on chip) max 80% and 90%, respectively. In addition to energy constraints, the electric vehicle cluster also needs to take into account power constraints. Charging and discharging power limit value P of controllable electric automobile cluster limit (t) may be represented by the formula:
P ev the charging and discharging power of a single controllable electric automobile in the secondary agent is regulated to be positive.
In FIG. 12E control Respectively sum E cmax 、E cmin A comparison is made to limit the frequency modulated power to within the controllable energy constraint. Current SOC satisfies SOC min ≤SOC<SOC max When the total energy storage meets E cmin (t)≤E control (t)≤E cmax (t). At this time, the corresponding value of K,<0,K>>0, the power of which model can be adjusted up or down with LFCsignal under such control. Assuming that the controlled vehicle gradually rises in SOC under the LFC signal and reaches SOC max Controllable electric automobile E at this time c (t) also reaches the upper limit E with increasing height cmax (t). Then kr=0, kz en route>And 0, the charge and discharge power constraint exists on the path of K1, and the electric automobile is limited to participate in frequency modulation and can only be discharged. Also when the SOC reaches the lower limit SOC min And when the electric automobile is in the frequency modulation, the electric automobile is limited to be charged only.
In summary, considering the constraint of automobile power and energy, the SOC of the electric automobile participating in frequency modulation can be limited in a certain range, so as to achieve the purpose of meeting the running requirement of a user.
Droop control strategy: the single-area power system is an isolated power system, and the section considers a control mode of the electric automobile and primary frequency modulation of the electric automobile, namely a sagging control link of a traditional generator set speed regulator is simulated. As described above, the power response transfer function of the electric vehicle is represented by the formula, k is the proportionality coefficient of the response frequency difference of the electric vehicle, and T B The delay time of the adjusting signal of the power electronic interface is very small; in addition, the current magnitude change ΔI in battery power regulation can be considered to be 0, where k EV ＝I b0 K, which is defined as the participation factor of the electric automobile in the frequency modulation of the electric power system, is obvious k EV Is a negative value.
It reflects the droop power response control approach. A sagging control schematic is shown in fig. 13. When the frequency deviation of the power grid does not exceed the set regulation dead zone, the electric automobile does not adjust the existing charging power, if the frequency deviation of the power system exceeds [ -f) 0 ,f 0 ]After the dead zone is adjusted, the charging power is adjusted, and when the negative shift of the power grid frequency is larger, the power supply mode is operated to supply electric energy to the power system so as to eliminate the frequency deviation caused by power unbalance.
Fig. 14 is a block diagram showing a power setting control of the electric vehicle participating in frequency modulation. In the droop control block, the frequency deviation signal Δf passes through a dead zone that prevents the electric vehicle from being too frequently power-adjusted and then passes through the proportional gain K p The method comprises the steps of carrying out a first treatment on the surface of the And finally, adding the output values of the droop control and the relation control with the initial test set point of the electric automobile power, and obtaining a new power distribution value after the battery power regulation response delay, wherein a new saturation limiter is needed to be added to limit the power within the allowable operation range of the agent.
In this model there are the following assumptions:
(1) Neglecting the SOC state of the power battery of the model electric automobile;
(2) The service life of the battery is not considered, and the electric automobile power battery is assumed to participate in primary frequency modulation without obvious loss.
(3) Acquiring a power grid frequency value of a charging and replacing power station in real time, calculating real-time frequency deviation, synchronously transmitting the real-time frequency deviation to each charging pile, and adjusting charging power by each charging pile foundation according to the real-time frequency deviation; when the power grid frequency exceeds a preset primary frequency modulation dead zone, the power adjustment direction and the power adjustment size are judged according to the real-time frequency deviation and the current control strategy that the charging pile participates in primary frequency modulation virtual sagging of the power grid, so that frequency modulation response of partial charging piles considering the SOC requirements of users is realized, and the starting time and the frequency response speed of the frequency modulation response are improved.
And constructing a charging noninductive guiding strategy and an effect prediction model which pass through the regional power grid multi-element regulation and control requirements, and excavating the available regulation capacity of the virtual power plant to realize the regulation and control of the user noninductive response power grid.
Those of ordinary skill in the art will appreciate that: the discussion of any of the embodiments above is merely exemplary and is not intended to suggest that the scope of the invention is limited to these examples; the technical features of the above embodiments or in the different embodiments may also be combined within the idea of the invention, the steps may be implemented in any order and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity.
The present invention is intended to embrace all such alternatives, modifications and variances which fall within the broad scope of the appended claims. Therefore, any omission, modification, equivalent replacement, improvement, etc. of the present invention should be included in the scope of the present invention.
Claims (10)
1. The V2G station fast frequency modulation response method based on ultra-short term frequency deviation prediction is characterized by comprising the following steps of:
step one: constructing an ultra-short-term load prediction model based on historical load data of a power grid;
step two: predicting load change conditions every 5 minutes in a rapid frequency modulation response period through an ultra-short-term load prediction model, wherein the prediction time is 5 minutes to 1 hour in the future;
step three: analyzing the relationship between the ultra-short-term load fluctuation rate and the system frequency based on the historical load data and the frequency statistical data of the power grid;
step four: calculating the ultra-short-term load fluctuation rate according to the ultra-short-term prediction result, and performing ultra-short-term frequency deviation prediction calculation;
step five: acquiring charge and discharge power, a vehicle SOC state, a start SOC, a stop SOC, a user demand mileage, a maximum power constraint of battery charge and discharge, a user lifting time and user participation response willingness information of each charging pile of a current V2G station;
step six: carrying out power distribution adjustment on each charging pile based on ultra-short-term frequency deviation prediction, reserving a phase of response power grid frequency modulation demand capacity according to a charging pile calculation result, and updating a charging pile frequency modulation virtual sagging control strategy;
step seven: and acquiring the power grid frequency value of the charging and replacing power station in real time, calculating the real-time frequency deviation, synchronously transmitting the real-time frequency deviation to each charging pile, and regulating the charging power by each charging pile foundation according to the real-time frequency deviation.
2. The method for fast frequency-modulated response of a V2G station based on ultra-short term frequency deviation prediction according to claim 1, wherein the ultra-short term load prediction model is based on a neural network algorithm, a time series algorithm, and a regression analysis algorithm.
3. The method for quickly adjusting frequency response of the V2G station based on ultra-short-term frequency deviation prediction according to claim 2, wherein the neural network algorithm is a cyclic neural network, the regression analysis algorithm is a multiple linear regression algorithm, and the time sequence algorithm comprises a two-way long-short-term memory neural network algorithm and a Facebook-propset algorithm.
4. The method for fast frequency modulation response of a V2G station based on ultra-short term frequency deviation prediction according to claim 3, wherein the multiple linear regression algorithm considers the model of time and weather independent variable factors as follows:
Load＝β 0 +β 1t +β 2w +β 3 3+...+β n×n +ε
where β0, β1,..βn is a model parameter, learning from the data is needed, ε is an error term, t is time, and w is weather.
5. A method for fast frequency response of a V2G station based on ultra-short term frequency deviation prediction according to claim 3, wherein the Facebook-propset algorithm decomposes the time sequence in its entirety into 3 parts: the effects of growing trend, seasonal trend and holiday, the specific function decomposition is as follows:
y(t)＝g(t)+s(t)+h(t)+∈ t
wherein g (t) represents a growth function for fitting non-periodic variations of the predicted values in the time series; s (t) is a period term for representing the periodic variation of the time series data, h (t) is a holiday term for representing the influence of holiday and holiday reasons on the time series data, and the specific correlation function is described as follows:
the growth function is classified into a linear growth and a nonlinear growth, the linear growth is realized by using a piecewise linear function, and the formula is:
g(t)＝(k+a(t)δ)·t+(m+a(t) T γ)
the nonlinear growth is implemented using a logic function, the formula of which is as follows:
where k represents the growth rate, δ is the adaptation rate, m is the offset parameter, C (t) For maximum bearing capacity, defining maximum value of growth, a (t) is vector, and gamma is offset parameter;
the period term is expressed using a fourier series, the formula of which is as follows:
wherein P represents the period of the time sequence, 2n represents the number of periods used in the model, and when n is too large, a complex seasonal function is fitted;
holiday items treat different holidays as independent models at different time points, and a time window is set up for each holiday model, and the holiday model is expressed as:
k＝(k 1 ，…k L ) T
wherein i represents holidays, di represents time t contained in a window period, ki represents influence of holidays on a prediction result, and 1 represents current value of Di.
6. The method for fast frequency modulation response of a V2G station based on ultra-short term frequency deviation prediction according to claim 5, wherein the two-way long-short term memory neural network algorithm comprises a gating mechanism, and each two-way long-short term memory neural network algorithm calculation unit comprises 3 control gates, namely an input gate, an output gate and a forget gate, and the specific algorithm is as follows:
forgetting door f t For forgetting the upper memory cell state c t-1 Information:
f t ＝σ(W f +U f h t-1 +b f )
wherein: w (W) f Is a weight matrix of forgetting gates; b f Is the forgetting door offset; sigma is a Sigmoid function;
candidate state of memory cellThe calculation of (a) is shown as the following formula, input gate i t Determining information retained in candidate states of the current cell:
i t ＝σ(W i +U i h t-1 +b i )
wherein: w (W) i i、W c Respectively represent inputDoor i t And candidate statesIs a weight matrix of (2); b i 、b c Is the corresponding offset;
through i t And f t Combining the previous time memory state c t-1 And a current time candidate memory stateTo update the memory cell state c at the current time t ：
Wherein Θ represents multiplication by element;
output door o t The method is mainly used for controlling the output of the state value of the memory cell:
o t ＝σ(W o x t +U o h t-1 +b 0 )
wherein: w (W) o Is an output gate o t Weight matrix of (a); b 0 Is the offset of the output gate;
the hidden layer output value ht is obtained through nonlinear calculation:
h t ＝o t Θtanh(c t )。
7. the method for fast frequency modulation response of a V2G station based on ultra-short term frequency deviation prediction according to claim 3, wherein a sliding average filtering method or a wavelet analysis method is adopted in the fourth step to extract the high frequency load fluctuation rate.
8. The method for quickly adjusting frequency response of the V2G station based on ultra-short-term frequency deviation prediction according to claim 2, further comprising an unequal weight combination algorithm prediction model, wherein the unequal weight combination algorithm prediction model is based on a neural network algorithm, a time sequence algorithm and a regression analysis algorithm, and the unequal weight solution is given to optimize to obtain a final prediction model.
9. The method for fast frequency response of a V2G station based on ultra-short term frequency deviation prediction according to claim 8, wherein the unequal weight combination algorithm prediction model specifically combines the results of a plurality of single prediction models, the plurality of single prediction models having different weights in the final prediction, and the weights of the prediction results are adjusted based on criteria for each single prediction model, the criteria including historical prediction performance of the model, and complexity of the model.
10. The method for fast frequency modulation response of a V2G station based on ultra-short term frequency deviation prediction according to claim 7, wherein the high frequency load fluctuation rate adopts a sliding average filtering method, specifically: firstly, a data sampling window is established in a RAM, M pieces of sampling data are stored, arithmetic average is carried out on the data in the sampling window, then the sampling window is shifted backwards by a certain length, each time new data is acquired, the earliest acquired data is lost, and average calculation is repeated;
assuming that the time length of the moving average sampling window is Mmin, the high frequency component PHt and the low frequency component PLt of the load at time t are calculated by the following formula:
wherein: PLt is the load low frequency component; PHt is the load high frequency component; pt is the load of tmin measured; t is a sampling time point; n is the total number of sampling points.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311061524.7A CN117559468A (en) | 2023-08-22 | 2023-08-22 | V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311061524.7A CN117559468A (en) | 2023-08-22 | 2023-08-22 | V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117559468A true CN117559468A (en) | 2024-02-13 |
Family
ID=89822181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311061524.7A Pending CN117559468A (en) | 2023-08-22 | 2023-08-22 | V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117559468A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117728475A (en) * | 2024-02-18 | 2024-03-19 | 天津大学 | Control method and system for intelligent charging pile supporting power grid frequency control |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105098852A (en) * | 2014-05-09 | 2015-11-25 | 株式会社日立制作所 | Electric automobile battery changing device and charge management method thereof |
| CN108321822A (en) * | 2018-02-28 | 2018-07-24 | 湖南大学 | A kind of short-term power grid frequency modulation control method and system based on energy-storage battery |
| CN110071579A (en) * | 2019-06-05 | 2019-07-30 | 国网北京市电力公司 | Power grid power supply based on ubiquitous electric power Internet of Things ensures and intelligent managing and control system |
| CN111980856A (en) * | 2020-08-17 | 2020-11-24 | 燕山大学 | Load prediction-based frequency modulation control method for energy storage type hydraulic wind generating set |
| CN113872228A (en) * | 2021-10-19 | 2021-12-31 | 广东电网有限责任公司 | Electric vehicle scheduling method and device applied to power grid peak shaving frequency modulation |
| CN114638440A (en) * | 2022-05-17 | 2022-06-17 | 国网湖北省电力有限公司经济技术研究院 | Charging load ultra-short-term prediction method based on charging pile utilization degree |
| CN115169703A (en) * | 2022-07-12 | 2022-10-11 | 广东电网有限责任公司广州供电局 | Short-term power load prediction method based on long-term and short-term memory network combination |
| CN116526511A (en) * | 2023-05-19 | 2023-08-01 | 东北电力大学 | Method for controlling load frequency of multi-source cooperative participation system |
-
2023
- 2023-08-22 CN CN202311061524.7A patent/CN117559468A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105098852A (en) * | 2014-05-09 | 2015-11-25 | 株式会社日立制作所 | Electric automobile battery changing device and charge management method thereof |
| CN108321822A (en) * | 2018-02-28 | 2018-07-24 | 湖南大学 | A kind of short-term power grid frequency modulation control method and system based on energy-storage battery |
| CN110071579A (en) * | 2019-06-05 | 2019-07-30 | 国网北京市电力公司 | Power grid power supply based on ubiquitous electric power Internet of Things ensures and intelligent managing and control system |
| CN111980856A (en) * | 2020-08-17 | 2020-11-24 | 燕山大学 | Load prediction-based frequency modulation control method for energy storage type hydraulic wind generating set |
| CN113872228A (en) * | 2021-10-19 | 2021-12-31 | 广东电网有限责任公司 | Electric vehicle scheduling method and device applied to power grid peak shaving frequency modulation |
| CN114638440A (en) * | 2022-05-17 | 2022-06-17 | 国网湖北省电力有限公司经济技术研究院 | Charging load ultra-short-term prediction method based on charging pile utilization degree |
| CN115169703A (en) * | 2022-07-12 | 2022-10-11 | 广东电网有限责任公司广州供电局 | Short-term power load prediction method based on long-term and short-term memory network combination |
| CN116526511A (en) * | 2023-05-19 | 2023-08-01 | 东北电力大学 | Method for controlling load frequency of multi-source cooperative participation system |
Non-Patent Citations (1)
| Title |
|---|
| XINLI SHI, ET AL: "A Novel Load Frequency Control Strategy Based on Model Predictive Control", 2015 IEEE POWER & ENERGY SOCIETY GENERAL MEETING, 31 October 2015 (2015-10-31), pages 1 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117728475A (en) * | 2024-02-18 | 2024-03-19 | 天津大学 | Control method and system for intelligent charging pile supporting power grid frequency control |
| CN117728475B (en) * | 2024-02-18 | 2024-04-19 | 天津大学 | Control method and system for intelligent charging pile supporting power grid frequency control |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chiş et al. | Reinforcement learning-based plug-in electric vehicle charging with forecasted price | |
| WO2021103625A1 (en) | Short-term vehicle speed condition real-time prediction method based on interaction between vehicle ahead and current vehicle | |
| CN110555990B (en) | Effective parking space-time resource prediction method based on LSTM neural network | |
| CN113515884A (en) | Distributed electric vehicle real-time optimization scheduling method, system, terminal and medium | |
| CN116001624A (en) | Ordered charging method for one-pile multi-connected electric automobile based on deep reinforcement learning | |
| CN103241130A (en) | Energy management method and system for electric bus charging and swap station | |
| Feng et al. | Load forecasting of electric vehicle charging station based on grey theory and neural network | |
| CN114997935B (en) | Electric vehicle charging and discharging strategy optimization method based on interior point strategy optimization | |
| CN117559468A (en) | V2G station rapid frequency modulation response method based on ultra-short term frequency deviation prediction | |
| CN112217195B (en) | Cloud energy storage charging and discharging strategy forming method based on GRU multi-step prediction technology | |
| CN113807589B (en) | Rolling optimization method and device for energy storage of operators based on model prediction | |
| CN114997631A (en) | Electric vehicle charging scheduling method, device, equipment and medium | |
| Mbuwir et al. | A hybrid policy gradient and rule-based control framework for electric vehicle charging | |
| CN113393339A (en) | Power distribution network short-term load prediction method based on deep neural network combination model | |
| CN114692965A (en) | Electric vehicle cluster schedulable potential prediction method and device based on digital-analog hybrid driving | |
| CN112348591A (en) | Ordered charging control method and device for electric automobile | |
| Zhang et al. | A safe reinforcement learning-based charging strategy for electric vehicles in residential microgrid | |
| CN112134272A (en) | Distribution network electric automobile load regulation and control method | |
| El-Azab et al. | Seasonal electric vehicle forecasting model based on machine learning and deep learning techniques | |
| CN111313449A (en) | Cluster electric vehicle power optimization management method based on machine learning | |
| CN117543581A (en) | Virtual power plant optimal scheduling method considering electric automobile demand response and application thereof | |
| CN116542137A (en) | Multi-agent reinforcement learning method for distributed resource cooperative scheduling | |
| Hong et al. | 6G based intelligent charging management for autonomous electric vehicles | |
| CN107154625B (en) | Electric car electric discharge electricity price based on fuzzy Bayesian learning negotiates method | |
| Tian et al. | Transfer learning based hybrid model for power demand prediction of large-scale electric vehicles |
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 |