CN117057523A - Power distribution network energy storage double-layer planning method based on load prediction - Google Patents
Power distribution network energy storage double-layer planning method based on load prediction Download PDFInfo
- Publication number
- CN117057523A CN117057523A CN202310730517.5A CN202310730517A CN117057523A CN 117057523 A CN117057523 A CN 117057523A CN 202310730517 A CN202310730517 A CN 202310730517A CN 117057523 A CN117057523 A CN 117057523A
- Authority
- CN
- China
- Prior art keywords
- distribution network
- power distribution
- energy storage
- load
- power
- 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
- 238000009826 distribution Methods 0.000 title claims abstract description 119
- 238000004146 energy storage Methods 0.000 title claims abstract description 107
- 238000000034 method Methods 0.000 title claims abstract description 58
- 239000011159 matrix material Substances 0.000 claims abstract description 23
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 21
- 238000005457 optimization Methods 0.000 claims abstract description 20
- 238000007621 cluster analysis Methods 0.000 claims abstract description 5
- 230000009466 transformation Effects 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 12
- 239000002245 particle Substances 0.000 claims description 9
- 230000005611 electricity Effects 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 6
- 238000009434 installation Methods 0.000 claims description 4
- 238000013459 approach Methods 0.000 claims description 3
- 230000008878 coupling Effects 0.000 claims description 3
- 238000010168 coupling process Methods 0.000 claims description 3
- 238000005859 coupling reaction Methods 0.000 claims description 3
- 238000007599 discharging Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 claims description 3
- 101150067055 minC gene Proteins 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000006870 function Effects 0.000 description 6
- 230000000875 corresponding effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000002457 bidirectional effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 2
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 239000001569 carbon dioxide Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000002803 fossil fuel Substances 0.000 description 1
- 239000005431 greenhouse gas Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- 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
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0283—Price estimation or determination
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- 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
- H02J3/007—Arrangements for selectively connecting the load or loads to one or several among a plurality of power lines or power sources
- H02J3/0075—Arrangements for selectively connecting the load or loads to one or several among a plurality of power lines or power sources for providing alternative feeding paths between load and source according to economic or energy efficiency considerations, e.g. economic dispatch
Abstract
The invention relates to a power distribution network energy storage double-layer planning method based on load prediction, which comprises the following steps: acquiring a daily load curve of the power distribution network; carrying out cluster analysis on the daily load curve based on k-means++ clusters, and dividing a daily load scene; extracting importance value fluctuation matrixes of input features respectively aiming at different daily load scenes; establishing a BILSTM short-term load prediction model, and correcting input characteristics of the model based on an importance value fluctuation matrix to obtain short-term load prediction results in various scenes; based on a short-term load prediction result, a power distribution network energy storage planning double-layer model is established, the upper layer aims at the minimum investment cost of energy storage years, the lower layer aims at the minimum daily operation cost of the power distribution network, a second-order cone relaxation transformation model is introduced to be non-convex constraint, and a double-layer optimization algorithm is adopted to solve the power distribution network energy storage planning double-layer model, so that the optimal economical efficiency of the power distribution network in energy storage planning is realized. Compared with the prior art, the method has the advantages of high prediction precision, improved voltage fluctuation and the like.
Description
Technical Field
The invention relates to the field of power distribution network planning, in particular to a power distribution network energy storage double-layer planning method based on load prediction.
Background
As the global demand for electricity continues to grow, fossil fuel-derived power generation necessarily increases carbon dioxide emissions and thereby exacerbates the greenhouse effect. While installing distributed energy sources (debs), such as fans (WT), photovoltaics (PV), in the distribution grid helps reduce greenhouse gas emissions. A concomitant problem is the high rejection rate due to uncertainty and intermittence of the debs. To address this problem, energy storage has been introduced to address the space-time imbalance between random energy generation and power demand during power distribution network planning. Therefore, it is necessary to study how to select the location and capacity of the installation of the energy storage power station in the distribution network. In addition, the uncertainty and randomness of the electrical load also affects the location and capacity of the energy storage plant. However, in the planning research of the existing energy storage power station, the focus is only on energy storage, and uncertainty factors of the power load are ignored, so that how to predict the power load is also important to the planning of the energy storage, and meanwhile, the power load prediction is beneficial to maintaining the dynamic balance of a power generation end and a power utilization end, so that a reliable basis can be provided for the comprehensive dispatching of a power grid, and the stable and efficient operation of a power system is ensured. Therefore, the short-term load prediction of the power system has important guiding significance for improving the economical efficiency and the safety stability of the operation of the power system while researching the energy storage.
At present, load prediction based on machine learning has become one of hot spots for research and application in the related art, and particularly, a deep recursive network (RNN) and a Long Short Term Memory (LSTM) network can simultaneously cope with timing and nonlinearity problems, so that a deep learning model is widely applied in load prediction. Chen Jinpeng et al propose "comprehensive energy System load prediction of Secondary Modal decomposition Combined DBiLSTMMLR", which adopts a bidirectional short-term memory (BILSTM) network for load prediction, and demonstrate that BILSTM has better expression ability for continuous time series compared with LSTM. Before using the deep learning model for short-term load prediction, factors affecting load change need to be selected as input features, including weather, date, historical load values, and the like. However, the method fails to deeply mine the load fluctuation rule, so that effective and comprehensive prediction information cannot be extracted from the load fluctuation, and the load prediction cannot be applied to energy storage planning.
Disclosure of Invention
The invention aims to provide a power distribution network energy storage double-layer planning method based on load prediction, which is characterized in that firstly, in order to extract effective and comprehensive prediction information from load fluctuation and further improve prediction accuracy, a mutual information (mutual information, MI) method and BILSTM short-term load prediction model considering characteristic importance value fluctuation are established based on k-means++ clustering; the energy storage planning double-layer model of the power distribution network is built on the basis of load prediction, the upper layer aims at the minimum investment cost of energy storage years, the lower layer aims at the minimum daily running cost of the power distribution network, and meanwhile, the non-convex constraint of the second order cone relaxation transformation model is introduced, so that the optimal economy of the power distribution network in the energy storage planning is realized.
The aim of the invention can be achieved by the following technical scheme:
a power distribution network energy storage double-layer planning method based on load prediction comprises the following steps:
step 1) acquiring a daily load curve of a power distribution network;
step 2) carrying out cluster analysis on the daily load curve based on k-means++ clusters, and dividing a daily load scene;
step 3) extracting importance value fluctuation matrixes of input features respectively aiming at different daily load scenes;
step 4) building a BILSTM short-term load prediction model, and correcting input characteristics of the BILSTM model by taking an importance value fluctuation matrix as a coefficient to obtain short-term load prediction results in various scenes;
and 5) establishing an energy storage planning double-layer model of the power distribution network based on a short-term load prediction result, wherein the upper layer aims at the minimum investment cost of energy storage years, the lower layer aims at the minimum daily operation cost of the power distribution network, a second-order cone relaxation transformation model is introduced into non-convex constraint, and the double-layer optimization algorithm is adopted to solve the energy storage planning double-layer model of the power distribution network, so that the optimal economy of the power distribution network in energy storage planning is realized.
The k-means++ clusters take meteorological features as input to subdivide daily load scenes.
And when the k-means++ is clustered, a contour coefficient method is adopted to perform clustering validity test on a clustering result, wherein a calculation formula of the contour coefficient method is as follows:
wherein: y (i) represents the distance of the daily load sequence from other load sequences in the cluster to which it belongs; x (i) represents the average distance of the daily load sequence to all load sequences within a cluster that does not contain it; z (i) represents a contour coefficient value, which is between [ -1,1], and when the value approaches 1 time, the clustering effect is relatively good.
The elements in the importance value fluctuation matrix are MI values, and the calculation formula of the MI values is as follows:
wherein: m (H, K) is the MI value between the random variables H and K; h is a HK (H, K) is a joint probability density function of random variables H and K, where H and K are elements in H and K, respectively; h is a H (h) An edge probability density function which is a random variable H; k (k) K (k) An edge probability density function which is a random variable K; the stronger the correlation between the two variables, the greater the MI value,when the two variables are independent of each other, the MI value is 0.
Said step 3) comprises the steps of:
step 3-1) determining an output load dataset: according to the daily load scene division result, dividing the power load data set into n groups according to the number of load sampling points in one day, and taking the n groups as an output data set P= [ P ] 1 ,P 2 ,…,P t ,…,P n ]Wherein P is t Load data output at the moment t;
step 3-2) determining an input feature data set: taking an input feature matrix E corresponding to the output load data set, i.e
Wherein, output P at time t t Corresponding input feature E t ＝[E t,1 ,E t,2 ,…,E t,m ] T ；
Step 3-3) calculating an importance value of the input feature at the time t: calculating the MI value of the input characteristic and the output load at the moment t, and carrying out normalization processing to obtain:
E t ＝[M(E t,1 ,P t ),M(E t,2 ,P t ),…,M(E t,m ,P t )] T
the value in the sequence is used as the importance value of different input characteristics at the time t;
step 3-4) repeating the step 3-3), circularly solving the importance values of the input features at different moments from 1 to n, and obtaining an importance value fluctuation matrix M of the input features changing along with time after the completion of the solution when t=n, namely:
the calculation formula of the BILSTM short-term load prediction model is as follows:
wherein:and->Load predicted values of the forward LSTM unit and the backward LSTM unit at the time t respectively; r is (r) t-1 The load predicted value at the time t-1; />And->Respectively representing LSTM forward and backward calculation processes; i.e t The input characteristic is the t moment; e, e t And f t Respectively outputting weights in the forward direction and the backward direction at the moment t; c t Optimizing parameters for bias at the time t; o (O) t And the load predicted value is finally output by the BILSTM unit at the moment t.
Said step 4) comprises the steps of:
step 4-1), determining a predicted day, inputting date characteristics and meteorological characteristics of the predicted day into a database, and preprocessing the database;
step 4-2) carrying out daily load scene division based on k-means++ clustering, and judging a daily load scene s to which a predicted day belongs;
step 4-3) determining input features of the daily load scene s, and extracting an importance value fluctuation matrix of the input features;
step 4-4) dynamically correcting input features based on the importance value fluctuation matrix;
step 4-5), constructing a BILSTM short-term load prediction model, and taking the dynamically corrected input characteristics as input to obtain a short-term load prediction result;
step 4-6) judging whether all the prediction days have completed short-term load prediction, if yes, completing prediction, otherwise, returning to the step 4-1).
In the energy storage planning double-layer model of the power distribution network, the upper layer model uses the energy storage annual investment cost C 1 The minimum target, expressed as:
wherein: r is annual rate; a is the system operation age; p (P) bat And Q bat Maximum power and capacity configured for energy storage, respectively; alpha 1 And alpha 2 The unit installation cost of the energy storage power and the capacity is respectively;
lower model uses the daily operation cost C of power distribution network 2 Minimum target, including the cost C of the power grid loss 3 Charge and discharge charge C of energy storage 4 Power distribution network electricity purchasing expense C 5 Expressed as:
minC 2 ＝C 3 +C 4 +C 5
wherein: p (P) loss. The power is lost for the power grid at the moment t;and->Charging and discharging power at t time for storing energy at the node i; />The power generated by the power grid at the node i at the moment t; gamma is the unit power grid loss cost; beta is the charge and discharge power cost of the energy storage unit; delta t The time-sharing electricity price of the power grid.
The energy storage planning double-layer model of the power distribution network comprises typical non-convex constraint, a second order cone relaxation technology is introduced, the energy storage site selection and volume fixing non-convex optimization problem of the power distribution network is converted into a convex optimization problem, and the specific process is as follows:
defining variables:
wherein: u (U) i.t And U j.t The voltage values of the nodes i and j of the power distribution network at the t period are respectively; θ ij. The voltage phase difference between the node i and the node j of the power distribution network in the period t is obtained.
Based on defined variables, converting the constraint of a power flow equation of the power distribution network, the constraint of the current carrying capacity of a line and the constraint of the voltage of a node into respectively:
wherein: p (P) ij. Active power of the branch i-j of the power distribution network in the period t; q (Q) ij. Absence of branch i-j for t-period distribution networkA power; g ij And B ij The conductance and susceptance values of the power distribution network branches i-j are respectively;the current-carrying capacity upper limit of the branch i-j of the power distribution network at the moment t; />And->The upper and lower limits of the square of the voltage of the node i of the power distribution network are respectively defined.
The second order cone variable needs to satisfy the following coupling relationship:
σ ij. +σ ji. ＝0
the second relaxation treatment of the non-convex terms present in the above formula is:
the resulting second order cone standard form is expressed as:
the method for solving the energy storage planning double-layer model of the power distribution network by adopting the double-layer optimization algorithm is characterized in that the upper-layer model is solved by adopting a particle swarm algorithm, the lower-layer model is solved by adopting a GUROBI solver, and the solving process comprises the following steps:
step 5-3-1) initializing various parameters of a power distribution network, wherein the iteration times k=0, and presetting the population number m and the total iteration times;
step 5-3-2), initially randomly generating rated capacity and initial positions of m energy storages by using a particle swarm algorithm, and transmitting parameters to a lower model;
step 5-3-3) updating the iteration number, k=k+1;
step 5-3-4), the power distribution network system sequentially receives initial values of m energy storage rated capacities and positions, solves the daily operation cost of the power distribution network by utilizing a GUROBI solver, and returns an optimized result to an upper model;
step 5-3-5) the energy storage device determines an optimal investment cost C according to the running cost of the distribution network in one day 1. ；
Step 5-3-6) generating new energy storage capacity value and position by using particle swarm algorithm, repeating step 5-3-4) and step 5-3-5), and calculating to obtain energy storage investment cost C 1 ′ . ；
Step 5-3-7) if C 1 ′ .k ≤C 1. C is then 1.k+1 ＝C 1 ′ .k Otherwise C 1.+1 ＝C 1.k ；
Step 5-3-8), if the energy storage investment cost and the daily operation cost of the power distribution network are converged, ending the solving process; if not, returning to the step 5-3-3).
Compared with the prior art, the invention has the following beneficial effects:
(1) The BILSTM short-term load prediction method corrected by the characteristic importance value fluctuation matrix can make up for the defects of the LSTM structure and improve the prediction precision.
(2) Under various daily load scenes, the prediction method provided by the invention has higher prediction precision, which indicates that the method is not limited to specific scenes and has good self-adaptability and stability.
(3) According to the invention, on the basis of considering load prediction, the energy storage of the power distribution network is planned, so that the daily operation cost of the power distribution network is reduced, the power flow distribution of the system can be improved, the network loss is further reduced, and the voltage fluctuation is improved, so that the method has more practical significance.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of a load clustering scenario and probability thereof in one embodiment;
FIG. 3 is a short-term load prediction flow chart of the present invention;
FIG. 4 is a diagram of the BILSTM model structure in one embodiment;
FIG. 5 is a graph of load prediction versus load prediction in different scenarios in one embodiment;
FIG. 6 is a schematic diagram of a power system 33 node structure in one embodiment;
FIG. 7 is a graph comparing voltage fluctuations in one embodiment.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.
The embodiment provides a power distribution network energy storage double-layer planning method based on load prediction, which is shown in fig. 1 and comprises the following steps:
step 1) acquiring a daily load curve of the power distribution network.
And 2) carrying out cluster analysis on the daily load curve based on k-means++ clusters, and dividing the daily load scene.
One of the characteristics of the power system load is the periodicity of the power system load, the development of short-term load prediction based on the periodic characteristics of the load is the basis for improving the prediction accuracy, and the periodicity of the load is reflected in the 24-hour daily overall change law and has a similar trend. In addition, the daily load scene is affected by various factors (date characteristics, temperature, humidity and the like) and can be classified into different types, and the importance degree of the input characteristics also shows obvious difference. For the characteristics, the embodiment firstly performs cluster analysis on the daily load curve based on a k-means++ algorithm.
The k-means++ is used for improving the k-means, and the k-means++ enables the distance between initial cluster centers to be as far as possible in the cluster initialization process, so that the problem that the cluster centers are trapped into local optimum is avoided, and therefore relatively better cluster centers are selected. The K-means++ algorithm effectively solves the problem of initial center selection of the K-means algorithm, but does not provide an effective solution for the selection of the number of clusters K. In order to determine the optimal K value, the embodiment adopts a contour coefficient method to perform clustering validity test on the clustering result. The calculation formula of the contour coefficient method is as follows:
wherein: y (i) represents the distance of the daily load sequence from other load sequences in the cluster to which it belongs; x (i) represents the average distance of the daily load sequence to all load sequences within a cluster that does not contain it; z (i) represents a contour coefficient value, which is between [ -1,1], and when the value approaches 1 time, the clustering effect is relatively good.
In this embodiment, the present invention is applicable to a variety of applications. The k-means++ clusters take meteorological features as input to subdivide daily load scenes. Meteorological characteristics include different types of data such as temperature, humidity, etc. As the temperature is a significant factor influencing daily load change, the temperature in one day is selected as the input of clustering, and a final daily load scene division result is obtained.
Fig. 2 is a schematic diagram of a load clustering scene and probability thereof in the present embodiment.
And 3) respectively extracting importance value fluctuation matrixes of the input features aiming at different daily load scenes.
The elements in the importance value fluctuation matrix are MI values. The importance of the input features is measured by an MI value method, the extracted MI value is used for representing the importance value of the input features, and the larger the MI value is, the larger the correlation between the input features and the load is, namely, the larger the importance value is. MI comes from the concept of entropy in the information theory, reflecting the correlation between any two random variables. The importance value of the input feature is characterized by an MI value between the input feature and the load. Therefore, on the basis of dividing different daily load scenes from the original database, importance value fluctuation matrixes of the input features are respectively extracted for the divided specific scenes.
The calculation formula of MI value is:
wherein: m (H, K) is the MI value between the random variables H and K; h is a HK (H, K) is a joint probability density function of random variables H and K, where H and K are elements in H and K, respectively; h is a H (h) An edge probability density function which is a random variable H; k (k) K (k) An edge probability density function which is a random variable K; the stronger the correlation between the two variables, the greater the MI value, which is 0 when the two variables are independent of each other.
Specifically, the extraction of the importance value fluctuation matrix of the input features comprises the following steps:
step 3-1) determining an output load dataset: according to the daily load scene division result, dividing the power load data set into n groups according to the number of load sampling points in one day, and taking the n groups as an output data set P= [ P ] 1 ,P 2 ,…,P t ,…,P n ]Wherein P is t And the load data is output at the time t.
Step 3-2) determining an input feature data set: taking an input feature matrix E corresponding to the output load data set, i.e
Wherein, output P at time t t Corresponding input feature E t ＝[E t,1 ,e t,2 ,…,e t,m ] T 。
Step 3-3) calculating an importance value of the input feature at the time t: calculating the MI value of the input characteristic and the output load at the moment t, and carrying out normalization processing to obtain:
E t ＝[M(E t,1 ,P t ),M(E t,2 ,P t ),…,M(E t,m ,P t )] T
the value in the sequence is taken as the time tImportance values of different input features, e.g. class 1 input features at time t, are M (E t,1 ,P t )。
Step 3-4) repeating the step 3-3), circularly solving the importance values of the input features at different moments from 1 to n, and obtaining an importance value fluctuation matrix M of the input features changing along with time after the completion of the solution when t=n, namely:
the M comprises importance information of input features under different dimensions. The importance value fluctuation of the same type of input features at different moments can be reflected through transverse time dimension comparison, and the importance difference of different types of input features at the same moment can be reflected through longitudinal type dimension comparison.
And 4) building a BILSTM short-term load prediction model, and correcting the input characteristics of the BILSTM model by taking the importance value fluctuation matrix as a coefficient to obtain short-term load prediction results in various scenes.
As shown in fig. 3, the short-term load prediction includes the steps of:
and 4-1) determining a predicted day, and inputting date characteristics and meteorological characteristics of the predicted day into a database, and preprocessing the database.
And 4-2) carrying out daily load scene division based on k-means++ clustering, and judging a daily load scene s to which a predicted day belongs.
Step 4-3) determining the input characteristics of the daily load scene s, and extracting an importance value fluctuation matrix of the input characteristics.
And 4-4) dynamically correcting the input characteristics of the BILSTM short-term load prediction model by taking the extracted importance value fluctuation matrix M as a coefficient in order to compensate for the time invariance of the weight sharing structure of the BILSTM model.
And 4-5) constructing a BILSTM short-term load prediction model, and taking the dynamically corrected input characteristics as input to obtain a short-term load prediction result.
In the short-term load prediction, the load value at the present time is correlated with both the information at the history time and the information at the future time. In this embodiment, the BILSTM taking bidirectional time information into consideration is selected as a bottom model of short-term load prediction, and the structure of the BILSTM is shown in fig. 4.
The calculation formula of the BILSTM short-term load prediction model is as follows:
wherein:and->Load predicted values of the forward LSTM unit and the backward LSTM unit at the time t respectively; r is (r) t-1 The load predicted value at the time t-1; />And->Respectively representing LSTM forward and backward calculation processes; i.e t The input characteristic is the t moment; e, e t And f t Respectively outputting weights in the forward direction and the backward direction at the moment t; c t Optimizing parameters for bias at the time t; o (o) t And the load predicted value is finally output by the BILSTM unit at the moment t.
Step 4-6) judging whether all the prediction days have completed short-term load prediction, if yes, completing prediction, otherwise, returning to the step 4-1).
And 5) establishing an energy storage planning double-layer model of the power distribution network based on a short-term load prediction result, wherein the upper layer aims at the minimum investment cost of energy storage years, the lower layer aims at the minimum daily operation cost of the power distribution network, a second-order cone relaxation transformation model is introduced into non-convex constraint, and the double-layer optimization algorithm is adopted to solve the energy storage planning double-layer model of the power distribution network, so that the optimal economy of the power distribution network in energy storage planning is realized.
Along with continuous promotion of intelligent power distribution network and active power distribution network construction, a large amount of power distribution network data information are interacted and deeply coupled, a traditional optimization algorithm regards a power distribution network energy storage power station as a fixed load node in a 'passive' mode, a system global optimal solution cannot be obtained, adaptability and portability are poor, and the method is obviously not applicable any more. Meanwhile, uncertainty and randomness factors of the power load also influence the position and capacity of the energy storage power station. In view of this, the embodiment combines the load prediction result to build a power distribution network energy storage site selection and volume determination double-layer optimization model, and further utilizes second order cone relaxation to convert the model non-convex constraint, so as to realize the rapid response of power distribution network optimization.
In the energy storage planning double-layer model of the power distribution network, the upper layer model uses the energy storage annual investment cost c 1 The minimum target, expressed as:
wherein: r is annual rate; a is the system operation age; p (P) bat And Q bat Maximum power and capacity configured for energy storage, respectively; alpha 1 And alpha 2 The unit installation cost of the energy storage power and the capacity is respectively.
Lower model uses the daily operation cost C of power distribution network 2 Minimum target, including the cost C of the power grid loss 3 Charge and discharge charge C of energy storage 4 Power distribution network electricity purchasing expense C 5 Expressed as:
minC 2 ＝C 3 +C 4 +C 5
wherein: p (P) loss. The power is lost for the power grid at the moment t;and->Charging and discharging power at t time for storing energy at the node i; />The power generated by the power grid at the node i at the moment t; gamma is the unit power grid loss cost; beta is the charge and discharge power cost of the energy storage unit; delta t The time-sharing electricity price of the power grid.
From the analysis, the built energy storage planning double-layer model (energy storage and location and volume-fixing model) of the power distribution network comprises typical non-convex constraints such as power flow constraints. The method has the advantages that the problem that the existing non-convex optimization problem cannot be solved by using mature commercial software, meanwhile, the heuristic algorithm has the defects of poor global searching capability, low precision and the like when solving the non-convex optimization problem, so that a second-order cone relaxation technology is introduced, the non-convex optimization problem of the power distribution network energy storage, site selection and volume determination is converted into the convex optimization problem, and the model solving difficulty is reduced. Taking an energy storage, location and volume selection model of a power distribution network as an example, a second order cone relaxation process of the model is introduced. Variables were first defined as follows:
wherein: u (U) i.t And U j.t Respectively t time period power distributionThe voltage values of the network nodes i, j; θ ij. The voltage phase difference between the node i and the node j of the power distribution network in the period t is obtained.
Based on defined variables, converting the constraint of a power flow equation of the power distribution network, the constraint of the current carrying capacity of a line and the constraint of the voltage of a node into respectively:
wherein: p (P) ij. Active power of the branch i-j of the power distribution network in the period t; q (Q) ij. Reactive power of the branch i-j of the power distribution network in the period t; g ij And B ij The conductance and susceptance values of the power distribution network branches i-j are respectively;the current-carrying capacity upper limit of the branch i-j of the power distribution network at the moment t; />And->The upper and lower limits of the square of the voltage of the node i of the power distribution network are respectively defined.
From the above, the second order cone variable needs to satisfy the following coupling relation:
σ ij. +σ ji. ＝0
since the above formula still has non-convex terms, it is further relaxed as:
further in the form of a second order cone standard can be expressed as:
according to the embodiment, a double-layer optimization algorithm is adopted to solve the energy storage planning double-layer model of the power distribution network, a particle swarm algorithm is adopted to solve the upper-layer model, a GUROBI solver is adopted to solve the lower-layer model, and the solving process comprises the following steps:
step 5-3-1) initializing various parameters of the power distribution network, wherein the iteration times k=0, the population m is set to be 10, and the total iteration times are 30.
Step 5-3-2), the rated capacity and the initial position of m energy storage are generated at random initially by using a particle swarm algorithm, and parameters are transmitted to a lower model.
Step 5-3-3) update the iteration number, k=k+1.
Step 5-3-4), the power distribution network system sequentially receives initial values of m energy storage rated capacities and positions, solves the daily operation cost of the power distribution network by utilizing a GUROBI solver, and returns an optimized result to an upper model.
Step 5-3-5) the energy storage device determines an optimal investment cost c according to the running cost of the distribution network in one day 1. 。
Step 5-3-6) generating new energy storage capacity value and position by using particle swarm algorithm, repeating step 5-3-4) and step 5-3-5), and calculating to obtain energy storage investment cost C 1 ′ . 。
Step 5-3-7) if C 1 ′ .k ≤C 1. C is then 1.k+1 ＝C 1 ′ .k Whether or notThen C 1.+1 ＝C 1.k 。
Step 5-3-8), if the energy storage investment cost and the daily operation cost of the power distribution network are converged, ending the solving process; if not, returning to the step 5-3-3).
In this embodiment, 4 different models, LSTM, BILSTM, MI-LSTM and MI-BILSTM, are selected for load prediction result comparison, and tables 1 and 5 show load prediction comparison under different scenes.
TABLE 1 comparison of prediction results for different models in different scenarios
As can be seen from Table 1, the average prediction accuracy of the LSTM model is lowest in the 4 models, the BILSTM model is higher than the LSTM model because the bi-directional information can be considered, the MI-LSTM model takes the importance value fluctuation of the input characteristics into consideration, the prediction accuracy is obviously improved compared with the LSTM model, and for the MI-BILSTM model, the prediction accuracy of the model under 4 different daily load scenes is better than that of other models, so that the importance value fluctuation matrix extracted by the MI method has more obvious improvement effect on the accuracy compared with other methods.
MAPE of the MI-BILST model under 1-4 scenes is 0.90%, 1.00%, 1.26%,1.84%, and average value of the MI-BILST model under all daily load scenes is 1.25%, and compared with average values of other 3 models, the MAPE is improved by 2.47%, 1.49% and 0.88%. The RMSE of the MI-BILSTM model in each scene is 14.91, 15.94, 19.60 and 28.88kW in sequence, and the average value is 19.61kW, and compared with the average value of the other 3 models, the RMSE is respectively reduced by 39.29, 23.37 and 13.29kW.
As shown in fig. 5, the prediction accuracy of the MI-BILSTM model is highest among the 4 models, as can be seen from fig. 5 for the different model result pairs. At the wave crest and the wave trough of the load curve, the model not only considers the bidirectional information flow, but also realizes the dynamic tracking of the fluctuation of the importance value of the input characteristic, and has better fitting effect on the real load curve.
In order to verify the effectiveness of the optimization method, the embodiment takes a 33-node power system structure as an example, and performs case verification, and the node structure is shown in fig. 6.
For energy storage planning, 4 case verifications were performed.
Scheme 1: considering load prediction, not considering energy storage planning, and not considering incorporating fans and photovoltaics;
scheme 2: considering incorporating fans, photovoltaics and load prediction, and not considering energy storage planning;
scheme 3: simultaneously, the combination of a fan, photovoltaic and energy storage is considered, but load prediction is not considered;
scheme 4: consider incorporating fans, photovoltaics and energy storage, as well as load prediction.
In order to track the importance value fluctuation of the input characteristics and further improve the prediction precision, an MI-BILSTM short-term load prediction method considering the feature importance value fluctuation is provided on the basis of k-means++ clustering, and then a power distribution network energy storage planning double-layer model is constructed by taking energy storage investment cost, power distribution network daily operation cost, network loss and the like as optimization targets based on load prediction.
The energy storage planning results obtained in this example are shown in table 2.
Table 2 comparison of index and configuration results
| Scheme for the production of a semiconductor device | 1 | 2 | 3 | 4 |
| Energy storage investment (Wanyuan) | 0 | 0 | 4403.2 | 4201.8 |
| Daily operation cost (Yuan) of distribution network | 37373.8 | 30462.3 | 26410.3 | 25632.8 |
| Distribution network electricity purchasing expense (Yuan) | 15402.1 | 11642.7 | 6830.3 | 6729.0 |
| Network loss (MW) | 5.48 | 4.75 | 2.53 | 2.45 |
| Energy storage capacity (MWh) | 0 | 0 | 1.53 | 1.46 |
| Node location | 0 | 0 | 17 | 26 |
The pair of voltage fluctuations obtained in this example is shown in fig. 7.
As can be seen from fig. 7, all the schemes have the voltage within a reasonable range, the voltage fluctuation of scheme 1 is the largest, scheme 2 is slightly improved compared with scheme 1, and scheme 3 is not greatly different from scheme 4, and is significantly improved compared with schemes 1 and 2. It is easy to see that energy storage planning is implemented on the power distribution network on the basis of load prediction, and the cost, voltage fluctuation and network loss of the distribution network purchase operation are all improved.
Through practical example verification, the following conclusion is obtained:
1) The MI-BILSTM short-term load prediction method adopting the characteristic importance value fluctuation matrix can make up for the defects of the LSTM structure and improve the prediction precision.
2) Under various daily load scenes, the proposed prediction method shows higher prediction precision. This means that the method is not limited to a specific scenario, and has good adaptability and stability.
3) On the basis of considering load prediction, the energy storage of the power distribution network is planned, so that the daily operation cost of the power distribution network is reduced, the power flow distribution of the system can be improved, the network loss is further reduced, the voltage fluctuation is improved, and the method has practical significance.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by a person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.
Claims (10)
1. The power distribution network energy storage double-layer planning method based on load prediction is characterized by comprising the following steps of:
step 1) acquiring a daily load curve of a power distribution network;
step 2) carrying out cluster analysis on the daily load curve based on k-means++ clusters, and dividing a daily load scene;
step 3) extracting importance value fluctuation matrixes of input features respectively aiming at different daily load scenes;
step 4) building a BILSTM short-term load prediction model, and correcting input characteristics of the BILSTM model by taking an importance value fluctuation matrix as a coefficient to obtain short-term load prediction results in various scenes;
and 5) establishing an energy storage planning double-layer model of the power distribution network based on a short-term load prediction result, wherein the upper layer aims at the minimum investment cost of energy storage years, the lower layer aims at the minimum daily operation cost of the power distribution network, a second-order cone relaxation transformation model is introduced into non-convex constraint, and the double-layer optimization algorithm is adopted to solve the energy storage planning double-layer model of the power distribution network, so that the optimal economy of the power distribution network in energy storage planning is realized.
2. The energy storage double-layer planning method for the power distribution network based on load prediction according to claim 1, wherein the k-means++ clusters take meteorological features as input to subdivide daily load scenes.
3. The power distribution network energy storage double-layer planning method based on load prediction according to claim 1, wherein when the k-means++ is clustered, a contour coefficient method is adopted to perform clustering validity test on a clustering result, and a calculation formula of the contour coefficient method is as follows:
wherein: y (i) represents the distance of the daily load sequence from other load sequences in the cluster to which it belongs; x (i) represents the average distance of the daily load sequence to all load sequences within a cluster that does not contain it; z (i) represents a contour coefficient value, which is between [ -1,1], and when the value approaches 1 time, the clustering effect is relatively good.
4. The power distribution network energy storage double-layer planning method based on load prediction according to claim 1, wherein elements in the importance value fluctuation matrix are MI values, and a calculation formula of the MI values is as follows:
wherein: m (H, K) is the MI value between the random variables H and K; h is a HK (H, K) is a joint probability density function of random variables H and K, where H and K are elements in H and K, respectively; h is a H (h) An edge probability density function which is a random variable H; k (k) K (k) An edge probability density function which is a random variable K; the stronger the correlation between the two variables, the greater the MI value, which is 0 when the two variables are independent of each other.
5. The method for energy storage double-layer planning of power distribution network based on load prediction according to claim 4, wherein the step 3) comprises the following steps:
step 3-1) determining an output load dataset: according to the daily load scene division result, dividing the power load data set into n groups according to the number of load sampling points in one day, and taking the n groups as an output data set P= [ P ] 1 ,P 2 ,…,P t ,…,P n ]Wherein P is t Load data output at the moment t;
step 3-2) determining an input feature data set: taking an input feature matrix E corresponding to the output load data set, i.e
Wherein, output P at time t t Corresponding input feature E t ＝[E t,1 ,E t,2 ,…,E t,m ] T ；
Step 3-3) calculating an importance value of the input feature at the time t: calculating the MI value of the input characteristic and the output load at the moment t, and carrying out normalization processing to obtain:
E t ＝[M(E t,1 ,P t ),M(E t,2 ,P t ),…,M(E t,m ,P t )] T
the value in the sequence is used as the importance value of different input characteristics at the time t;
step 3-4) repeating the step 3-3), circularly solving the importance values of the input features at different moments from 1 to n, and obtaining an importance value fluctuation matrix M of the input features changing along with time after the completion of the solution when t=n, namely:
6. the energy storage double-layer planning method for the power distribution network based on load prediction according to claim 1, wherein the calculation formula of the BILSTM short-term load prediction model is as follows:
wherein:and->Load predicted values of the forward LSTM unit and the backward LSTM unit at the time t respectively; r is (r) t-1 The load predicted value at the time t-1; />And->Respectively representing LSTM forward and backward calculation processes; i.e t The input characteristic is the t moment; e, e t And f t Respectively outputting weights in the forward direction and the backward direction at the moment t; c t Optimizing parameters for bias at the time t; o (O) t And the load predicted value is finally output by the BILSTM unit at the moment t.
7. The method for energy storage double-layer planning of a power distribution network based on load prediction according to claim 1, wherein the step 4) comprises the following steps:
step 4-1), determining a predicted day, inputting date characteristics and meteorological characteristics of the predicted day into a database, and preprocessing the database;
step 4-2) carrying out daily load scene division based on k-means++ clustering, and judging a daily load scene s to which a predicted day belongs;
step 4-3) determining input features of the daily load scene s, and extracting an importance value fluctuation matrix of the input features;
step 4-4) dynamically correcting input features based on the importance value fluctuation matrix;
step 4-5), constructing a BILSTM short-term load prediction model, and taking the dynamically corrected input characteristics as input to obtain a short-term load prediction result;
step 4-6) judging whether all the prediction days have completed short-term load prediction, if yes, completing prediction, otherwise, returning to the step 4-1).
8. The method for planning energy storage double layers of a power distribution network based on load prediction according to claim 1, wherein in the power distribution network energy storage planning double layer model, an upper layer model is used for energy storage annual investment cost C 1 The minimum target, expressed as:
wherein: r is annual rate; a is the system operation age; p (P) bat And Q bat Maximum power and capacity configured for energy storage, respectively; alpha 1 And alpha 2 The unit installation cost of the energy storage power and the capacity is respectively;
lower model uses the daily operation cost C of power distribution network 2 Minimum target, including the cost C of the power grid loss 3 Charge and discharge charge C of energy storage 4 Power distribution network electricity purchasing expense C 5 Expressed as:
minC 2 ＝C 3 +C 4 +C 5
wherein: p (P) loss.t The power is lost for the power grid at the moment t;and->Charging and discharging power at t time for storing energy at the node i;the power generated by the power grid at the node i at the moment t; gamma is the unit power grid loss cost; beta is the charge and discharge power cost of the energy storage unit; delta t The time-sharing electricity price of the power grid.
9. The power distribution network energy storage double-layer planning method based on load prediction according to claim 1, wherein the power distribution network energy storage planning double-layer model comprises typical non-convex constraint, a second order cone relaxation technology is introduced, and the power distribution network energy storage site selection and volume determination non-convex optimization problem is converted into a convex optimization problem, and the specific process is as follows:
defining variables:
wherein: u (U) i.t And U j.t The voltage values of the nodes i and j of the power distribution network at the t period are respectively; θ ij. The voltage phase difference between the node i and the node j of the power distribution network in the t period is set;
based on defined variables, converting the constraint of a power flow equation of the power distribution network, the constraint of the current carrying capacity of a line and the constraint of the voltage of a node into respectively:
wherein: p (P) ij. Active power of the branch i-j of the power distribution network in the period t; q (Q) ij. Reactive power of the branch i-j of the power distribution network in the period t; g ij And B ij The conductance and susceptance values of the power distribution network branches i-j are respectively;the current-carrying capacity upper limit of the branch i-j of the power distribution network at the moment t; />And->The upper limit and the lower limit of the square voltage of the node i of the power distribution network are respectively;
the second order cone variable needs to satisfy the following coupling relationship:
σ ij. +σ ji. ＝0
the second relaxation treatment of the non-convex terms present in the above formula is:
the resulting second order cone standard form is expressed as:
10. the method for planning the energy storage double layers of the power distribution network based on load prediction according to claim 1, wherein the method for solving the energy storage planning double layers of the power distribution network by adopting a double-layer optimization algorithm is characterized in that a particle swarm algorithm is adopted for solving an upper layer model, a GUROBI solver is adopted for solving a lower layer model, and the solving process comprises the following steps:
step 5-3-1) initializing various parameters of a power distribution network, wherein the iteration times k=0, and presetting the population number m and the total iteration times;
step 5-3-2), initially randomly generating rated capacity and initial positions of m energy storages by using a particle swarm algorithm, and transmitting parameters to a lower model;
step 5-3-3) updating the iteration number, k=k+1;
step 5-3-4), the power distribution network system sequentially receives initial values of m energy storage rated capacities and positions, solves the daily operation cost of the power distribution network by utilizing a GUROBI solver, and returns an optimized result to an upper model;
step 5-3-5) the energy storage device determines an optimal investment cost C according to the running cost of the distribution network in one day 1. ；
Step 5-3-6) generating new energy storage capacity value and position by using particle swarm algorithm, repeating step 5-3-4) and step 5-3-5), and calculating to obtain energy storage investment cost C 1 ′ . ；
Step 5-3-7) if C 1 ′ .k ≤C 1. C is then 1.k+1 ＝C 1 ′ .k Otherwise C 1.+1 ＝C 1.k ；
Step 5-3-8), if the energy storage investment cost and the daily operation cost of the power distribution network are converged, ending the solving process; if not, returning to the step 5-3-3).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310730517.5A CN117057523A (en) | 2023-06-19 | 2023-06-19 | Power distribution network energy storage double-layer planning method based on load prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310730517.5A CN117057523A (en) | 2023-06-19 | 2023-06-19 | Power distribution network energy storage double-layer planning method based on load prediction |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117057523A true CN117057523A (en) | 2023-11-14 |
Family
ID=88665151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310730517.5A Pending CN117057523A (en) | 2023-06-19 | 2023-06-19 | Power distribution network energy storage double-layer planning method based on load prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117057523A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117353359A (en) * | 2023-12-05 | 2024-01-05 | 国网浙江省电力有限公司宁波供电公司 | Battery combined energy storage and power supply method and system |
-
2023
- 2023-06-19 CN CN202310730517.5A patent/CN117057523A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117353359A (en) * | 2023-12-05 | 2024-01-05 | 国网浙江省电力有限公司宁波供电公司 | Battery combined energy storage and power supply method and system |
| CN117353359B (en) * | 2023-12-05 | 2024-04-12 | 国网浙江省电力有限公司宁波供电公司 | Battery combined energy storage and power supply method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yan et al. | Stochastic multi-scenario optimization for a hybrid combined cooling, heating and power system considering multi-criteria | |
| CN107688879B (en) | Active power distribution network distributed power supply planning method considering source-load matching degree | |
| CN110266038B (en) | Distributed coordination regulation and control method for multiple virtual power plants | |
| CN110852565B (en) | Grid frame planning method considering different functional attributes | |
| CN109636056A (en) | A kind of multiple-energy-source microgrid decentralization Optimization Scheduling based on multi-agent Technology | |
| CN114595868A (en) | Source network and storage collaborative planning method and system for comprehensive energy system | |
| CN117057523A (en) | Power distribution network energy storage double-layer planning method based on load prediction | |
| Hu et al. | Short-term photovoltaic power prediction based on similar days and improved SOA-DBN model | |
| CN115587668A (en) | Distributed collaborative optimization scheduling method for multi-park integrated energy system | |
| CN110460038A (en) | It is a kind of to be related to more scene method for expansion planning of power transmission network of new-energy grid-connected | |
| Liu et al. | Two-layer multiple scenario optimization framework for integrated energy system based on optimal energy contribution ratio strategy | |
| CN114971048A (en) | Zero-carbon park comprehensive energy planning simulation method and system and computer equipment | |
| CN113469412B (en) | Real-time operation strategy optimization method and system for comprehensive energy system | |
| Luo et al. | Two‐stage robust optimal scheduling of wind power‐photovoltaic‐thermal power‐pumped storage combined system | |
| CN114117925A (en) | Comprehensive energy system planning method considering high-proportion distributed photovoltaic consumption | |
| CN113991742B (en) | Distributed photovoltaic double-layer collaborative optimization investment decision-making method for power distribution network | |
| CN116995719A (en) | Active power distribution network double-layer expansion planning method considering wind-solar-load uncertainty | |
| CN115758763A (en) | Multi-energy flow system optimal configuration method and system considering source load uncertainty | |
| CN115759589A (en) | Power distribution network data-driven robust extension planning method containing high-proportion new energy | |
| CN111311032B (en) | Micro-grid system capacity optimal configuration method based on sector radar map model | |
| CN112365134A (en) | Energy network random planning method based on point estimation method probability multi-energy flow | |
| Sheng et al. | Probabilistic Production Simulation for Industrial Integrated Energy System with Power and Steam Loads | |
| Chen et al. | Short-term PV power forecasting based on time-phased and error correction | |
| CN110598925A (en) | Energy storage in-trading market decision optimization method based on double-Q learning algorithm | |
| CN117094745B (en) | Comprehensive energy system optimization control method and device based on IGDT-utility entropy |
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 |