CN114757107A - Gravity energy storage power distribution method based on load prediction model - Google Patents
Gravity energy storage power distribution method based on load prediction model Download PDFInfo
- Publication number
- CN114757107A CN114757107A CN202210499867.0A CN202210499867A CN114757107A CN 114757107 A CN114757107 A CN 114757107A CN 202210499867 A CN202210499867 A CN 202210499867A CN 114757107 A CN114757107 A CN 114757107A
- Authority
- CN
- China
- Prior art keywords
- energy storage
- gravity energy
- storage system
- value
- prediction model
- 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
- 238000004146 energy storage Methods 0.000 title claims abstract description 125
- 230000005484 gravity Effects 0.000 title claims abstract description 125
- 238000009826 distribution Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 34
- 239000002245 particle Substances 0.000 claims abstract description 36
- 230000008901 benefit Effects 0.000 claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 30
- 238000013527 convolutional neural network Methods 0.000 claims description 17
- 238000004220 aggregation Methods 0.000 claims description 8
- 230000002776 aggregation Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 3
- 238000010248 power generation Methods 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 5
- 238000012549 training Methods 0.000 description 5
- 238000005457 optimization Methods 0.000 description 4
- 230000003213 activating effect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000013256 Gubra-Amylin NASH model Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 239000013256 coordination polymer Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- General Health & Medical Sciences (AREA)
- Geometry (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Public Health (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Water Supply & Treatment (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- General Business, Economics & Management (AREA)
- Medical Informatics (AREA)
Abstract
A gravity energy storage power distribution method based on a load prediction model relates to a gravity energy storage power distribution technology and aims to solve the problem that the existing load prediction is low in accuracy rate and unreasonable in power distribution of gravity energy storage. The invention establishes a load prediction model based on a GAN network; predicting the load of the gravity energy storage system by using the load, and outputting a prediction result; calculating a peak regulation demand value and a frequency modulation demand value based on the prediction result; obtaining a value of a peak-modulation participation factor and a value of a frequency-modulation participation factor by combining constraint conditions of a gravity energy storage system and adopting a particle swarm algorithm; adopting peak regulation participation factors and frequency modulation participation factors to jointly construct a power distribution model with the maximum economic benefit of the gravity energy storage system as a target; and performing power distribution on the gravity energy storage system by using a power distribution model. The gravity energy storage power distribution system has the beneficial effect that the power distribution of gravity energy storage is more reasonable.
Description
Technical Field
The invention relates to a gravity energy storage power distribution technology.
Background
One of the basic requirements for stable operation of an electric power system is that power is balanced in real time; along with the change of the load characteristics of the power system, the peak-to-valley difference of the power system is gradually increased, and the peak regulation problem is prominent day by day; meanwhile, the short-time active power between the power supply and the load is unbalanced, so that the system frequency fluctuation always influences the safe and stable operation of the power grid; the gravity energy storage technology is used as a novel energy storage system, has high geographical environment adaptability, and is expected to become an effective means for peak regulation and frequency regulation of a power grid in the future; the load prediction of the power system is the basis for implementing various user-oriented applications, and the accurate load prediction makes a reasonable production plan for the power system, so that resource waste is avoided, the safe and reliable operation of a power grid is ensured, and the important effect of improving the economic benefit is achieved; load prediction can be divided into long-term, medium-term, short-term and ultra-short-term load prediction according to the length of the prediction period; wherein, the load change rule in long term and medium term tends to be stable, and the research thereof is mature; short-term load changes are random, the prediction difficulty is high, and the method is always concerned by researchers; at the end of the 20 th century, modeling and predicting load prediction by adopting a statistical method; the methods are constructed aiming at the linear relation, influence of factors such as climate, date type and the like on short-term load prediction is ignored, and the prediction accuracy is low; therefore, the peak-shaving and frequency-modulation capacities of the power system are limited, and the power of the gravity energy storage cannot be reasonably distributed.
Disclosure of Invention
The invention aims to solve the problem that the power distribution of gravity energy storage is unreasonable due to low load prediction accuracy in the prior art, and provides a gravity energy storage power distribution method based on a load prediction model.
The invention relates to a gravity energy storage power distribution method based on a load prediction model, which comprises the following steps of:
step one, establishing a load prediction model based on a GAN network;
predicting the load of the gravity energy storage system by using the load prediction model established in the step one, and outputting a prediction result of the load of the gravity energy storage system;
step three, calculating a peak regulation demand value and a frequency modulation demand value of the gravity energy storage system based on the load prediction result of the gravity energy storage system output in the step two;
step four, calculating constraint conditions of the gravity energy storage system;
combining the constraint conditions of the gravity energy storage system calculated in the step four, and adopting a particle swarm optimization to optimally solve the peak regulation demand value to obtain a value of a peak regulation participation factor; simultaneously, performing optimized solution on the frequency modulation required value by adopting a particle swarm optimization algorithm to obtain a value of a frequency modulation participation factor;
step six, adopting the peak regulation participation factors and the frequency modulation participation factors obtained in the step five to jointly construct a power distribution model with the maximum economic benefit of the gravity energy storage system as a target;
And step seven, performing power distribution on the gravity energy storage system by using the power distribution model constructed in the step six.
The beneficial effects of the invention are: the power distribution method comprehensively considers the cost and the benefit of the gravity energy storage system participating in peak regulation and frequency modulation, combines the capacity limit of the gravity energy storage system, provides a technical and economic model of the gravity energy storage system participating in frequency regulation and peak regulation, solves the peak regulation demand value and the frequency regulation demand value by introducing a particle swarm algorithm, and realizes reasonable distribution of the power and the capacity of the gravity energy storage system participating in auxiliary service of the power system.
Drawings
Fig. 1 is a flowchart of a gravity energy storage power distribution method based on a load prediction model according to a first embodiment;
FIG. 2 is a schematic diagram of a network structure of a GAN network according to an embodiment;
FIG. 3 is a block diagram of a load prediction model according to one embodiment;
FIG. 4 is a diagram illustrating a generator according to a first embodiment;
FIG. 5 is a diagram illustrating a structure of a discriminator according to a first embodiment;
FIG. 6 is a flowchart of a method for obtaining peak shaving participation factors in a first embodiment;
fig. 7 is a graph comparing an optimal power allocation with a conventional allocation scheme in the first embodiment.
Detailed Description
The first specific implementation way is as follows: the present embodiment is described with reference to fig. 1 to fig. 7, and the method for distributing gravity stored energy power based on a load prediction model in the present embodiment includes the following steps:
step one, establishing a load prediction model based on a GAN network;
predicting the load of the gravity energy storage system by using the load prediction model established in the step one, and outputting a prediction result of the load of the gravity energy storage system;
step three, calculating a peak regulation demand value and a frequency modulation demand value of the gravity energy storage system based on the load prediction result of the gravity energy storage system output in the step two;
step four, calculating constraint conditions of the gravity energy storage system;
combining the constraint conditions of the gravity energy storage system calculated in the step four, and adopting a particle swarm optimization to optimally solve the peak regulation demand value to obtain a value of a peak regulation participation factor; simultaneously, performing optimized solution on the frequency modulation required value by adopting a particle swarm algorithm to obtain a value of a frequency modulation participation factor;
step six, adopting the peak regulation participation factor and the frequency modulation participation factor obtained in the step five to jointly construct a power distribution model with the maximum economic benefit of the gravity energy storage system as a target;
And step seven, performing power distribution on the gravity energy storage system by using the power distribution model constructed in the step six.
In the present embodiment, the load prediction model in the first step includes a generator and a discriminator;
a generator for learning the distribution of the real samples r and generating new samples; the real sample r is load data of the gravity energy storage system obtained through actual measurement;
and the discriminator is used for discriminating whether the input sample comes from the generator or not.
In the embodiment, after the generator learns the data implicit deep relation and reaches the balance, theoretically, the prediction result of the load of the gravity energy storage system output by the load prediction model approaches to the real data infinitely.
In this embodiment, the specific process of establishing the GAN network-based load prediction model in the first step is as follows: inputting the random noise n and the condition value c into a generator to generate a new sample F (n | c), and inputting the generated sample F (n | c) and the real sample r into a discriminator together with the condition c for discrimination; the discrimination result output by the discriminator is respectively fed back to the generator and the discriminator in the form of a loss function, and the generator and the discriminator revise the parameters thereof according to the feedback result; the condition value c is a load influence factor of the gravity energy storage system; the random noise n is a random variable conforming to Gaussian distribution;
The loss function of the generator in the load prediction model is defined as: l is a radical of an alcoholFThe loss function of the arbiter in the load prediction model is defined as: l is a radical of an alcoholD；
Loss function L of the generatorFIs of the form shown in equation (1):
LF＝-En,c(D(F(n|c)|c)) (1)
wherein, En,cExpressing the expected values of the distribution of random noise n and the condition value c; f meterShown is the output data of the generator; d represents output data of the discriminator;
loss function L of the discriminatorDIs of the form shown in equation (2):
LD＝-Er,c(D(r|c))+En,c(D(F(n|c)|c)) (2)
wherein E isr,cA distribution expectation value representing the real sample r and the condition value c;
the generator wants to increase the output value of F (n | c), and the discriminator wants to decrease the output value of F (n | c) and increase the output value of the measured data r; in the process of load prediction, a generator is used for generating a predicted value which is as close to actually measured load data as possible, noise n and a load influence factor c are spliced and then input into the generator, and predicted load data F (n | c) are output through the generator; the judger not only needs to judge the similarity between the predicted load data F (n | c) and the actually measured load data r, but also needs to judge the fit degree between the predicted load data F (n | c) and the load influence factor c; the loss function L of the load prediction model is thus a binary minimum maximum game with conditional probabilities.
The loss function L of the load prediction model is a binary minimum maximum game containing conditional probability and is defined
minmaxL＝Er,c(lnD(r|c))+En,c(ln(1-D(F(n|c)|c))) (3)
Wherein min max L represents a binary minimum maximum game value of the loss function L;
meanwhile, the load prediction model uses an L1 norm as a loss function, so that the obtained prediction result is more accurate; the loss function for the L1 norm is:
LL1＝Er,c,n(||r-F(n|c)||) (4)
wherein L isL1A loss function that is the norm of L1; er,c,nRepresenting the expected values of the distribution of the real sample r, the condition value c and the random noise n;
in fig. 2, r is a real sample and is actually measured load data; c is a condition value, historical load data and other influence factors; n is random noise; f (n | c) is the generated predicted load data; l is a loss function and is used as feedback of a load prediction model; inputting the random noise n and the condition value c into a generator to generate a sample F (n | c), and inputting the generated sample F (n | c) and the real sample r into a discriminator together with the condition c for discrimination; the judgment result of the discriminator is fed back to the generator and the discriminator in the form of a loss function, and the generator and the discriminator revise the parameters thereof according to the feedback result so as to improve the generation capability and the judgment capability of each; in order to further improve the accuracy of the short-term load prediction of the load prediction model, the characteristic value deviation between the predicted data and the real data generated by a hidden layer measurement generator of the discriminator is utilized; therefore, the network of the whole load prediction model is continuously subjected to iterative optimization, and the purpose of enabling the prediction result of the generator to be more accurate is achieved.
In the embodiment, a block diagram structure of a load prediction model based on a GAN network is shown in fig. 3, which includes three parts, namely, data set construction, model training and prediction result; the data set construction mainly refers to the steps of using historical load data and other influence factors to construct a training and testing data set, and performing data preprocessing, so that input and output variables of a load prediction model are determined; the load influence factors are condition data, including historical load data, climate data and date type data; actually measured load data is target data of the model; in the process of training a load prediction model, a GAN combined characteristic loss function is adopted for short-term load prediction; the probability of true data output by the discriminator; the generator and the discriminator optimize the self weight by continuously carrying out confrontation training, so that the whole load prediction model is optimal; finally, load prediction is carried out by using the trained prediction model, and prediction results are compared; when the load prediction model constructed in the embodiment is used, the contents of each group of test data sets are sequentially input into the generator, the daily prediction results are output and are respectively compared with the daily real values, and the prediction precision of the load prediction model based on the GAN network is obtained.
In the actual training, when the GAN faces data with more characteristic quantities, not only is the network not easy to stably train, but also various problems such as slow convergence rate, mode collapse, and quality of generated samples to be improved exist; in order to solve the problems, a Convolutional Neural Network (CNN) is introduced to construct the internal structures of a generator and a discriminator; the CNN not only has strong capability of extracting characteristics of multiple hidden layers, but also can share convolution kernels, has no pressure on high-dimensional data processing, and can improve the stability and convergence speed of the GAN and the quality of data generated by a generator by introducing the CNN; the two-dimensional convolution model is beneficial to extracting characteristics of the model, and a prediction model is constructed by adopting the structure of the two-dimensional convolution model, so that the overall generalization capability and robustness of the model can be enhanced; a method for calculating characteristic deviation by using a hidden layer of a partial discriminator is adopted to optimize a network structure, so that the prediction precision can be further improved.
In this embodiment, the generator comprises a 3-layer convolutional neural network;
the 3 layers of convolutional neural networks are respectively as follows: an input layer, a convolutional layer C1, and a convolutional layer C2;
the input layer consists of 32 convolution kernels of 5 × 5, the convolution layer C1 consists of 64 convolution kernels of 5 × 5, the convolution layer C2 consists of 1 convolution kernel of 5 × 5, and the sliding step size of each layer of the convolutional neural network is 2.
In order to enable a network to learn a more appropriate space sampling method independently, the structure of the generator adopts step convolution instead of using space pooling in CNN, and batch standardization operation is adopted among layers to accelerate convergence and slow down overfitting, so that gradient propagation layers are deeper; and adopting ReLU activating functions in other layers except the output layer, and adopting tanh activating functions in the output layer to finally generate the prediction data.
In this embodiment, the discriminator comprises a 3-layer convolutional neural network, the hidden layer of which uses a LeakyReLU function as an activation function, and the second and third hidden layers apply a feature loss function; and using a full connection and sigmoid activation function to judge true and false, and mapping the result between (0, 1);
the 3 layers of convolutional neural networks are respectively a first convolutional layer, a second convolutional layer and a third convolutional layer;
the first convolutional layer is composed of 32 convolution kernels of 5 × 5, the second convolutional layer is composed of 64 convolution kernels of 5 × 5, the third convolutional layer is composed of 128 convolution kernels of 5 × 5, and the step size of each convolutional neural network is 2.
In this embodiment, the specific formula for calculating the peak regulation requirement value of the gravity energy storage system in step three is as follows:
PD＝Ppeak-Pvalley (5)
Wherein, PDRepresenting the peak shaver requirement value, P, of the gravity energy storage systempeakPredicted load peak of the day, P, representing the output of the load prediction modelvalleyRepresenting the current day's predicted load trough output by the load prediction model.
In this embodiment, the specific formula for calculating the frequency modulation requirement value of the gravity energy storage system in the third step is as follows:
PF＝ΔPL-Pplan-Pline (6)
wherein, PFRepresenting the frequency modulation requirement value of the gravity energy storage system; delta PLRepresenting the load change value of the gravity energy storage system; pplanRepresenting a power generation planning frequency value of a frequency modulation unit of the gravity energy storage system; plineA tie-line adjustment planning frequency value representing the gravity energy storage system.
In this embodiment, the constraint conditions of the gravity energy storage system in step four are specifically:
PG1＝αPD≤PG，PG1≥0 (7)
PG2＝βPF≤PG，PG2≥0 (8)
wherein, PGIs the power capacity of the gravity energy storage system; pG1The capacity of the power which can participate in system peak regulation and is distributed for the gravity energy storage system; pG2The capacity of the power which can participate in system frequency modulation and is distributed for the gravity energy storage system; alpha is the value of the peak regulation participation factor; beta is the value of the frequency modulation participation factor.
In this embodiment, the specific method for obtaining the peak regulation participation factor in the step five is as follows:
fifthly, initializing the peak regulation demand value of the gravity energy storage system as an original population to generate an initial population;
Step two, carrying out economic benefit calculation on the initial generation population generated in the step one to determine the economic benefits of the gravity energy storage systems of different individual particles;
fifthly, selecting the position of the individual particle when the economic benefit of the gravity energy storage system is the maximum to obtain the optimal point of the individual particle and the global optimal point;
fifthly, updating the speed, the position, the aggregation degree and the weight of the individual particles;
fifthly, judging whether the updated individual particle speed, position, aggregation degree and weight meet the constraint conditions of the gravity energy storage system; if the constraint condition is met, executing a fifth step and a sixth step; otherwise, returning to execute the step five;
fifthly, outputting the positions of the updated individual particles, and taking the positions of the updated individual particles as peak regulation participation factors;
meanwhile, the concrete method for obtaining the frequency modulation participation factor in the step five is as follows:
step I, initializing a frequency modulation demand value of a gravity energy storage system as an original population to generate an initial generation population;
step II, calculating the economic benefit of the gravity energy storage system of the primary population generated in the step I, and determining the economic benefit of the gravity energy storage system of different individual particles;
Step III, selecting the position of the individual particle when the economic benefit of the gravity energy storage system is the maximum to obtain an individual particle optimal point and a global optimal point;
step IV, updating the speed, the position, the aggregation degree and the weight of the individual particles;
step V, judging whether the speed, the position, the aggregation degree and the weight of the updated individual particles meet the constraint conditions of the gravity energy storage system; if the constraint condition is met, executing a fifth step and a sixth step; otherwise, returning to execute the step two;
and VI, outputting the position of the updated individual particle, and taking the position of the updated individual particle as a frequency modulation participation factor.
In this embodiment, the expression of the power distribution model constructed in the sixth step is:
PG＝PG1+PG2＝αPD+βPF (9)
wherein, PGIs the power capacity of the gravity energy storage system; pG1The capacity of the power which can participate in system peak regulation and is distributed for the gravity energy storage system; pG2The capacity of the power which can participate in system frequency modulation and is distributed for the gravity energy storage system; alpha is the value of the peak regulation participation factor; beta is the value of the frequency modulation participation factor.
In the embodiment, the peak regulation requirement and the frequency modulation requirement of the system are considered, the capacity of the gravity energy storage system is fully utilized, the peak regulation participation factor and the frequency modulation participation factor of the gravity energy storage system are introduced, and the limited power of the gravity energy storage system in one operating working day is distributed;
And (3) adopting an optimal economic benefit-based distribution scheme, namely constructing a power distribution model with the maximum economic benefit of the gravity energy storage system as a target on the basis of considering the peak shaving, frequency modulation cost and benefit of the gravity energy storage system participating in the system, and ensuring the scheduling economy of the gravity energy storage system to the maximum extent.
(1) Penalty cost for insufficient scheduling:
because the peak shaving power and the frequency modulation power capacity of the gravity energy storage system are distributed, the phenomenon that the peak shaving power and the frequency modulation capacity of the system are insufficient is caused, and a penalty coefficient is introduced.
Clack＝cEElack+ΣcPPlack (10)
Wherein: c. CEPunishment cost for unit with insufficient peak regulation electric quantity; c. CPPunishment cost for units with insufficient frequency modulation power; elackThe power quantity which is insufficient for the participation of the distributed gravity energy storage system in peak shaving; plack PlackIs the fractional power of the frequency modulation.
(2) Peak modulation yield and frequency modulation yield:
the frequency modulation benefit of the gravity energy storage system comes from providing paid frequency modulation service for the power system;
IF＝Σi1PG2+i2EF (11)
wherein: i.e. i1The unit compensation gain of the frequency modulation power is provided for the gravity energy storage system; i.e. i2The unit benefit of the frequency modulation electric quantity is obtained; eFIs frequency modulated electric quantity.
The peak shaving benefit of the gravity energy storage system comes from peak-valley electricity price benefit and peak shaving subsidy benefit.
ID＝i3ED+i4ηED (12)
Wherein: i.e. i3Supplementing the unit electric quantity of peak regulation; i.e. i 4The comprehensive price difference of the power grid is obtained; eDPeak regulation electric quantity; eta is energy conversion efficiency;
therefore, the economic and technical model of the gravity energy storage system is as follows:
Itotal＝IF+ID-Clack (13)
wherein, ItotalIs the economic benefit of the gravity energy storage system.
Simulation verification:
taking power grid data of a certain area as an example, short-term load prediction is carried out, and economic distribution is carried out on the power of the gravity energy storage system so as to verify the effectiveness of the power distribution method of the embodiment.
The results of the prediction by GAN model are compared with the actual values as shown in table 1.
TABLE 1 comparison of load prediction results with actual values
Unit: MW
As can be seen from table 1, the load prediction model according to the present embodiment can consider more influencing factors by enhancing the processing capability of the high-dimensional data, thereby further improving the prediction accuracy.
Based on the prediction result, optimal power economy distribution is carried out on several gravity energy storage systems in the local area in the simulation system, and the optimal power economy distribution is compared with the yield of capacity equal proportion distribution of the conventional scheme, and the result is shown in fig. 7; through comparative analysis, the benefits obtained by adopting the power distribution method of the embodiment are higher than those obtained by adopting a conventional scheme, and the benefit difference between the two schemes gradually increases with the increase of time, so that the effectiveness and the economy of the power distribution method of the embodiment in power distribution of the gravity energy storage system are shown.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.
Claims (10)
1. A gravity energy storage power distribution method based on a load prediction model is characterized by comprising the following steps:
step one, establishing a load prediction model based on a GAN network;
predicting the load of the gravity energy storage system by using the load prediction model established in the step one, and outputting a prediction result of the load of the gravity energy storage system;
thirdly, calculating a peak regulation demand value and a frequency modulation demand value of the gravity energy storage system based on the load prediction result of the gravity energy storage system output in the second step;
step four, calculating constraint conditions of the gravity energy storage system;
combining the constraint conditions of the gravity energy storage system calculated in the step four, and performing optimal solution on the peak regulation required value by adopting a particle swarm algorithm to obtain a value of a peak regulation participation factor; simultaneously, performing optimized solution on the frequency modulation required value by adopting a particle swarm algorithm to obtain a value of a frequency modulation participation factor;
Step six, adopting the peak regulation participation factors and the frequency modulation participation factors obtained in the step five to jointly construct a power distribution model with the maximum economic benefit of the gravity energy storage system as a target;
and step seven, performing power distribution on the gravity energy storage system by using the power distribution model constructed in the step six.
2. The gravity energy storage power distribution method based on the load prediction model as claimed in claim 1, wherein the load prediction model in the first step comprises a generator and a discriminator;
a generator for learning the distribution of the real samples r and generating new samples; the real sample r is load data of the gravity energy storage system obtained through actual measurement;
a discriminator for discriminating whether the input sample is from the generator.
3. The gravity energy storage power distribution method based on the load prediction model according to claim 2, wherein the specific process of establishing the load prediction model based on the GAN network in the first step is as follows: inputting the random noise n and the condition value c into a generator to generate a new sample F (n | c), and inputting the generated sample F (n | c) and the real sample r into a discriminator together with the condition c for discrimination; the discrimination result output by the discriminator is respectively fed back to the generator and the discriminator in the form of a loss function, and the generator and the discriminator revise the parameters thereof according to the feedback result; the condition value c is a load influence factor of the gravity energy storage system; the random noise n is a random variable conforming to Gaussian distribution;
The loss function of the generator in the load prediction model is defined as: l is a radical of an alcoholFThe loss function of the discriminator in the load prediction model is defined as: l is a radical of an alcoholD；
Loss function L of the generatorFIs of the form shown in equation (1):
LF＝-En,c(D(F(n|c)|c)) (1)
wherein E isn,cExpressing the expected values of the distribution of random noise n and the condition value c; fRepresenting the output data of the generator; d represents output data of the discriminator;
loss function L of the discriminatorDIs of the form shown in equation (2):
LD＝-Er,c(D(r|c))+En,c(D(F(n|c)|c)) (2)
wherein E isr,cRepresenting the expected values of the distribution of the real sample r and the condition value c;
the loss function L of the load prediction model is a binary minimum maximum game containing conditional probability and is defined
minmaxL＝Er,c(lnD(r|c))+En,c(ln(1-D(F(n|c)|c))) (3)
Wherein minmaxL represents a binary minimum maximum game value of the loss function L;
meanwhile, the load prediction model uses an L1 norm as a loss function; the loss function for the L1 norm is:
LL1＝Er,c,n(||r-F(n|c)||) (4)
wherein L isL1A loss function that is the norm of L1; er,c,nRepresenting the expected values of the distribution for the true sample r, the condition value c, and the random noise n.
4. The gravity energy storage power distribution method based on the load prediction model according to claim 2, wherein the generator comprises a 3-layer convolutional neural network;
The 3 layers of convolutional neural networks are respectively as follows: an input layer, a convolutional layer C1, and a convolutional layer C2;
the input layer consists of 32 convolution kernels of 5 × 5, the convolutional layer C1 consists of 64 convolution kernels of 5 × 5, the convolutional layer C2 consists of 1 convolution kernel of 5 × 5, and the step size of sliding of each convolutional neural network is 2.
5. The gravity energy storage power distribution method based on the load prediction model according to claim 2, wherein the discriminator comprises a 3-layer convolutional neural network; the 3 layers of convolutional neural networks are respectively a first convolutional layer, a second convolutional layer and a third convolutional layer;
the first convolutional layer is composed of 32 convolution kernels of 5 × 5, the second convolutional layer is composed of 64 convolution kernels of 5 × 5, the third convolutional layer is composed of 128 convolution kernels of 5 × 5, and the step size of each convolutional neural network is 2.
6. The gravity energy storage power distribution method based on the load prediction model as claimed in claim 1, wherein the specific formula for calculating the peak shaver requirement value of the gravity energy storage system in the third step is as follows:
PD＝Ppeak-Pvalley (5)
wherein, PDRepresenting the peak shaver requirement value, P, of the gravity energy storage systempeakPredicted load peak of the day, P, representing the output of the load prediction modelvalleyRepresenting the current day's predicted load trough output by the load prediction model.
7. The gravity energy storage power distribution method based on the load prediction model according to claim 1, wherein the specific formula for calculating the frequency modulation requirement value of the gravity energy storage system in the third step is as follows:
PF＝ΔPL-Pplan-Pline (6)
wherein, PFRepresenting the frequency modulation requirement value of the gravity energy storage system; delta PLRepresenting the load change value of the gravity energy storage system; p isplanRepresenting a power generation planning frequency value of a frequency modulation unit of the gravity energy storage system; p islineA tie-line adjustment planning frequency value representing the gravity energy storage system.
8. The gravity energy storage power distribution method based on the load prediction model according to claim 1, wherein the constraint conditions of the gravity energy storage system in the fourth step are specifically:
PG1＝αPD≤PG，PG1≥0 (7)
PG2＝βPF≤PG，PG2≥0 (8)
wherein, PGIs the power capacity of the gravity energy storage system; pG1The capacity of the power which can participate in system peak regulation and is distributed for the gravity energy storage system; pG2The capacity of the power which can participate in system frequency modulation and is distributed for the gravity energy storage system; alpha is the value of the peak regulation participation factor; beta is the value of the frequency modulation participation factor.
9. The gravity energy storage power distribution method based on the load prediction model according to claim 1, wherein the specific method for obtaining the peak regulation participation factor in the step five is as follows:
fifthly, initializing the peak regulation demand value of the gravity energy storage system as an original population to generate an initial population;
Step two, calculating the economic benefits of the gravity energy storage system of the primary population generated in the step one to determine the economic benefits of the gravity energy storage system of different individual particles;
fifthly, selecting the position of the individual particle when the economic benefit of the gravity energy storage system is the maximum to obtain an individual particle optimal point and a global optimal point;
fifthly, updating the speed, the position, the aggregation degree and the weight of the individual particles;
fifthly, judging whether the updated individual particle speed, position, aggregation degree and weight meet the constraint conditions of the gravity energy storage system; if the constraint condition is met, executing a fifth step and a sixth step; otherwise, returning to execute the step two;
fifthly, outputting the positions of the updated individual particles, and taking the positions of the updated individual particles as peak regulation participation factors;
meanwhile, the concrete method for obtaining the frequency modulation participation factor in the step five is as follows:
step I, initializing a frequency modulation demand value of a gravity energy storage system as an original population to generate an initial generation population;
step II, calculating the economic benefit of the gravity energy storage system of the primary population generated in the step I, and determining the economic benefit of the gravity energy storage system of different individual particles;
Step III, selecting the position of the individual particle when the economic benefit of the gravity energy storage system is the maximum to obtain an individual particle optimal point and a global optimal point;
step IV, updating the speed, the position, the aggregation degree and the weight of the individual particles;
step V, judging whether the speed, the position, the aggregation degree and the weight of the updated individual particles meet the constraint conditions of the gravity energy storage system; if the constraint condition is met, executing a fifth step and a sixth step; otherwise, returning to execute the step two;
and VI, outputting the position of the updated individual particle, and taking the position of the updated individual particle as a frequency modulation participation factor.
10. The gravity energy storage power distribution method based on the load prediction model according to claim 1, wherein the expression of the power distribution model constructed in the sixth step is as follows:
PG＝PG1+PG2＝αPD+βPF (9)
wherein, PGIs the power capacity of the gravity energy storage system; pG1The capacity of the power which can participate in system peak regulation and is distributed for the gravity energy storage system; pG2The capacity of the power which can participate in system frequency modulation and is distributed for the gravity energy storage system; alpha is the value of the peak regulation participation factor; beta is the value of the frequency modulation participation factor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210499867.0A CN114757107A (en) | 2022-05-09 | 2022-05-09 | Gravity energy storage power distribution method based on load prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210499867.0A CN114757107A (en) | 2022-05-09 | 2022-05-09 | Gravity energy storage power distribution method based on load prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114757107A true CN114757107A (en) | 2022-07-15 |
Family
ID=82334715
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210499867.0A Pending CN114757107A (en) | 2022-05-09 | 2022-05-09 | Gravity energy storage power distribution method based on load prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114757107A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401604A (en) * | 2020-02-17 | 2020-07-10 | 国网新疆电力有限公司经济技术研究院 | Power system load power prediction method and energy storage power station power distribution method |
| CN118336931A (en) * | 2024-06-13 | 2024-07-12 | 国网甘肃省电力公司电力科学研究院 | Tower type gravity energy storage system and method |
-
2022
- 2022-05-09 CN CN202210499867.0A patent/CN114757107A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111401604A (en) * | 2020-02-17 | 2020-07-10 | 国网新疆电力有限公司经济技术研究院 | Power system load power prediction method and energy storage power station power distribution method |
| CN111401604B (en) * | 2020-02-17 | 2023-07-07 | 国网新疆电力有限公司经济技术研究院 | Power system load power prediction method and energy storage power station power distribution method |
| CN118336931A (en) * | 2024-06-13 | 2024-07-12 | 国网甘肃省电力公司电力科学研究院 | Tower type gravity energy storage system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111401604B (en) | Power system load power prediction method and energy storage power station power distribution method | |
| CN114757107A (en) | Gravity energy storage power distribution method based on load prediction model | |
| CN110674999A (en) | Cell load prediction method based on improved clustering and long-short term memory deep learning | |
| CN109558975A (en) | A kind of integrated approach of a variety of prediction results of electric load probability density | |
| CN113158572A (en) | Short-term load prediction method and device | |
| Zhang et al. | Achieving wind power and photovoltaic power prediction: An intelligent prediction system based on a deep learning approach | |
| Wang et al. | Application of IEHO–BP neural network in forecasting building cooling and heating load | |
| CN113869616A (en) | Self-adaptive residential user load prediction method | |
| CN114139940A (en) | Generalized demand side resource network load interaction level assessment method based on combined empowerment-cloud model | |
| Jiang et al. | Industrial ultra-short-term load forecasting with data completion | |
| CN115986728A (en) | Power grid situation prediction method considering uncertainty factors and terminal | |
| CN115983710A (en) | High-proportion new energy access electric power system infrastructure project decision method and system | |
| CN115409645A (en) | Comprehensive energy system energy management method based on improved deep reinforcement learning | |
| CN115456312A (en) | Short-term power load prediction method and system based on octyl geometric modal decomposition | |
| CN116826710A (en) | Peak clipping strategy recommendation method and device based on load prediction and storage medium | |
| CN114239372B (en) | Multi-objective unit maintenance double-layer optimization method and system considering unit combination | |
| Gang et al. | Dynamic economic emission dispatch with wind power based on improved multi‐objective brain storm optimisation algorithm | |
| Liao et al. | A novel stacked generalization ensemble-based hybrid PSVM-PMLP-MLR model for energy consumption prediction of copper foil electrolytic preparation | |
| CN116542882B (en) | Photovoltaic power generation smoothing method, system and storage medium | |
| CN117078462A (en) | Decision method for participation of load aggregator in electric energy and frequency modulation market | |
| CN116307199A (en) | Load prediction method for power system | |
| Tang et al. | A novel hybrid FA-Based LSSVR learning paradigm for hydropower consumption forecasting | |
| CN112785022B (en) | Method and system for excavating electric energy substitution potential | |
| Liu et al. | Short-term power load forecasting via recurrent neural network with similar day selection | |
| Zhang et al. | Capacity Optimization of Hybrid Energy Storage System Based on Improved Golden Eagle Optimization |
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 |