Distributed resource uncertain transaction strategy considering typhoon space-time evolution

Abstract

The invention relates to a distributed resource uncertain transaction strategy considering typhoon space-time evolution, and the strategy is characterized in that the strategy comprises the following steps: S1, building a typhoon disaster space-time evolution model; and S2, calculating a power distribution network line time-varying failure rate and a fan time-varying failure rate based on a typhoon disaster time-space evolution model. S3, training the graph attention network based on the time-varying fault rate of the nodes to obtain a power distribution network fault rate prediction model of the nodes based on graph attention, and obtaining an actual fault rate prediction value based on the power distribution network fault rate prediction model; s4, acquiring a wind power historical output data set, and constructing an uncertainty fuzzy set based on a Wasserstein distance; and S5, constructing a distributed resource uncertain transaction strategy model considering typhoon space-time evolution, and solving the model to obtain a transaction strategy. Compared with the prior art, the method has the advantages that the fault prediction result of the power distribution network under the typhoon condition is improved, and then distribution and transaction of power resources are optimized.

Description

Technical Field

[0001] The present invention relates to the field of distributed resource allocation, and in particular to a distributed resource uncertainty trading strategy taking into account the temporal and spatial evolution of typhoons. Background Art

[0002] Distributed resources, especially wind power generation, photovoltaic power generation and energy storage systems, play an important role in typhoon weather. They not only improve the disaster resistance of the power grid, but also provide key support for post-disaster recovery. During typhoons, traditional centralized power generation facilities may be severely affected, while distributed systems may continue to supply power to local areas, effectively reducing the overall load pressure of the distribution network. However, extreme weather events such as typhoons often lead to low-probability-high-loss events, which makes it particularly important to accurately predict the failure rate of the distribution network and achieve optimal scheduling of distributed resources. Existing trading strategies do not take into account the uncertainty of wind power output, resulting in an unbalanced resource allocation strategy and poor grid stability. In addition, in existing methods, the prediction of the failure rate of the distribution network is still inaccurate, which makes it difficult for trading strategies to obtain accurate prediction results of distribution network failures and difficult to analyze extreme weather conditions, especially the allocation and trading of appropriate power resources under typhoon conditions. Summary of the invention

[0003] The purpose of the present invention is to improve the prediction results of distribution network failures under typhoon conditions, and further optimize the allocation and trading of power resources, and to provide a distributed resource uncertain trading strategy that takes into account the spatiotemporal evolution of typhoons.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] A distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons, the strategy includes the following steps:

[0006] S1. Establish a model for the temporal and spatial evolution of typhoon disasters;

[0007] S2. Calculate the time-varying failure rate of nodes based on the model of the spatiotemporal evolution of typhoon disasters, wherein the time-varying failure rate of nodes includes the time-varying failure rate of distribution network lines and the time-varying failure rate of wind turbines.

[0008] S3. Train the graph attention network based on the time-varying failure rate of the node to obtain the distribution network failure rate prediction model of the node based on the graph attention, and obtain the actual failure rate prediction value based on the distribution network failure rate prediction model. During the training process, the pinball loss function is used.

[0009] S4, obtain the historical wind power output data set and construct the uncertainty fuzzy set based on Wasserstein distance;

[0010] S5. Based on the uncertainty fuzzy sets and the actual failure rate prediction value, a distributed resource uncertain trading strategy model considering the spatiotemporal evolution of typhoons is constructed, and the trading strategy is obtained by solving the model.

[0011] Furthermore, the distribution network line time-varying fault rate P L for:

[0012]

[0013] Among them, where: μ 1 , δ 1 are the mean and standard deviation of the tensile strength of the wire respectively; 1 Indicates the breaking stress of the conductor; σ g is the stress on the conductor cross section.

[0014] Furthermore, the time-varying failure rate of the wind turbine for:

[0015]

[0016] in, It indicates the probability of accidental failure of the fan under normal operating conditions, which is obtained through the historical data of the fan operation failure rate; v r 、v cutin 、v cutoff are the rated wind speed, cut-in wind speed and cut-out wind speed of the fan respectively; v w It is the real-time wind speed at the geographical location of the wind farm.

[0017] Furthermore, the specific steps of training the graph attention network based on the node-based time-varying failure rate are:

[0018] The time-varying failure rate sequence of the distribution network line or the time-varying failure rate sequence of the wind turbine corresponding to each of the n nodes is used as the input of the graph attention network;

[0019] Calculate the correlation between nodes and normalize them to get the attention coefficient;

[0020] Get the output feature vector of the node based on the attention coefficient;

[0021] The middle layer of the graph attention network concatenates the output feature vectors of the nodes, and the last layer of the graph attention network averages the output feature vectors of the nodes. The output of the last layer passes through the prediction head to obtain the model prediction value;

[0022] The pinball loss function is calculated based on the model prediction value, and the graph attention network is trained based on the pinball loss function.

[0023] Furthermore, the pinball loss function is:

[0024]

[0025]

[0026] Among them, y * is the actual value, y q is the model prediction value, and q is the cumulative probability of the quantile corresponding to the pinball loss function.

[0027] Furthermore, the specific steps of S4 are:

[0028] Obtain the historical wind power output data set and convert the data set into the empirical distribution P of the wind and solar forecast error K As an estimate of the true distribution P;

[0029] Calculate the empirical distribution P K The Wasserstein distance from the true distribution P;

[0030] Construct the empirical distribution P K is the center, ε(K) is the radius of the Wasserstein ball 0 .

[0031] Furthermore, the spherical fuzzy set D 0 for:

[0032]

[0033] in, represents the total probability distribution on Ξ, W(P K ,P) represents the empirical distribution P K The Wasserstein distance between the actual distribution P and ε(K) is the empirical distribution P. K The fuzzy set D is the center 0 The radius of the Wasserstein sphere, ε(K) satisfies

[0034]

[0035] Among them, β is the confidence level; K represents the sample size; and D is the coefficient.

[0036] Furthermore, the objective function of the distributed resource uncertain transaction strategy model is:

[0037]

[0038] C grid (t) = C buy (t)-C sell (t)

[0039] Cbuy (t) = c buy (t)P buy (t)

[0040] C sell (t) = c sell (t)P sell (t)

[0041]

[0042] C wt (t) = o wt P wt (t)

[0043] C pv (t) = o pv P pv (t)

[0044] C bat (t) = o ch P ch (t)

[0045] Among them, f 1 represents the objective function, C grid The cost of the load aggregator interacting with the main network, including the cost of purchasing electricity C buy and electricity sales revenue C sell , where c buy 、c sell are the electricity purchase and sale prices, P buy , P sell They are respectively the power of electricity purchased and the power of electricity sold; C MT is the cost of gas turbine power generation, where a i , b i 、c i is the fuel coefficient; P MT represents the power generated by the gas turbine, C wt , C pv , C bat are the operation and maintenance costs of wind turbines, photovoltaic units and battery energy storage modules, among which o wt , o pv , o ch are the unit power operation and maintenance costs of wind turbines, photovoltaic units and batteries, P wt , P pv , P ch They are the power generation capacity of wind turbine, photovoltaic unit and battery respectively.

[0046] Furthermore, the constraints of the distributed resource uncertain trading strategy model include: AC power flow equation constraints, line power constraints, operation constraints, distributed blue-power opportunity constraints, gas turbine output power and ramp constraints, power main grid output power constraints, energy storage equipment constraints and power balance constraints.

[0047] Furthermore, the distribution of robust chance constraints is:

[0048]

[0049] Among them, the spherical fuzzy set D 0 , L represents the line set, z i,j,t represents the fault status of the line at time t, z i,j,t Indicates the fault status of the line at time t, usually a binary variable, indicating whether the line is faulty (1 indicates fault, 0 indicates normal). represents the active power of the wind farm at time t, Indicates the lower and upper limits of the active power of the wind farm at time t. P Is the wind farm in the fuzzy set D 0 The expectation of the next output, ∈, is used to represent the allowed risk probability. In the distributed robust chance constraint, it defines the upper limit of the probability that the constraint condition is not satisfied, ∈ is the violation probability;

[0050] The power balance constraint is:

[0051] P grid (t)+P wt (t)(1-E wt (t))+P MT (t)(1-E line (t))+P bat (t)=Load(t)

[0052] Among them, E wt 、E line are the time-varying failure rate of wind turbines and the time-varying failure rate of distribution network lines in the actual failure rate prediction value under typhoon scenarios. Line failure will cause the gas turbine output to be unable to be sent to the power grid, which will affect the operating environment of the gas turbine and the stability of the power grid, thereby indirectly affecting the operation and scheduling of the gas turbine. grid (t) is the main power grid power, P bat (t) represents the energy storage battery power, and Load(t) represents the load power.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention uses the calculated failure rates of different time periods to form a time-varying failure rate sequence as the input of the distribution network node. The powerful feature learning mechanism and the ability to capture complex nonlinear relationships of the graph attention method are used to construct a distribution network failure model to predict the failure rate of the distribution network. In model training, the pinball loss function is selected as the loss function of the model to obtain an accurate prediction of the distribution network failure rate. At the same time, in typhoon weather, the uncertainty of wind power output is also a factor that cannot be ignored. In order to deal with this uncertainty, when solving the trading strategy model, the distributed robust optimization method is introduced, and a spherical fuzzy set is constructed based on the historical wind power output data set. Combined with the predicted distribution network failure rate, the allocation and trading of power resources can be optimized under extreme weather conditions to ensure the stable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a structural schematic diagram of the present invention;

[0056] Figure 2 Schematic diagram of the graph attention network model;

[0057] Figure 3 This is the economic scheduling result diagram of distributed resources;

[0058] Figure 4 Distributed resource output diagram before and after uncertainty processing. DETAILED DESCRIPTION

[0059] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0060] The present invention proposes a distributed resource uncertainty trading strategy that takes into account the temporal and spatial evolution of typhoons. The strategy process is as follows: Figure 1 As shown in the figure. First, a typhoon weather spatiotemporal evolution model is established based on meteorological information. A typhoon wind field model is established to achieve quantitative analysis of the distribution network failure rate. A two-stage graph attention network is used to predict the distribution network failure rate. For the uncertainty of wind power output, a distributed robust optimization method is adopted. With the goal of minimizing the economic dispatch cost of the distribution network during typhoons, a distributed resource trading model taking into account uncertainty is established to improve the flexibility of the distribution network.

[0061] The present invention comprises the following steps:

[0062] S1. Establish a model for the temporal and spatial evolution of typhoon disasters;

[0063] S2. Calculate the time-varying failure rate of nodes based on the model of the spatiotemporal evolution of typhoon disasters, wherein the time-varying failure rate of nodes includes the time-varying failure rate of distribution network lines and the time-varying failure rate of wind turbines.

[0064] S3. Train the graph attention network based on the time-varying failure rate of the node to obtain the distribution network failure rate prediction model of the node based on the graph attention, and obtain the actual failure rate prediction value based on the distribution network failure rate prediction model. During the training process, the pinball loss function is used.

[0065] S4, obtain the historical wind power output data set and construct the uncertainty fuzzy set based on Wasserstein distance;

[0066] S5. Based on the uncertainty fuzzy sets and the actual failure rate prediction value, a distributed resource uncertain trading strategy model considering the spatiotemporal evolution of typhoons is constructed, and the trading strategy is obtained by solving the model.

[0067] Considering the impact of typhoons, a rectangular coordinate system based on the basic parameters of the distribution network and the geographical location information is first constructed for the study area. By integrating the meteorological data of the Geographic Information System (GIS), a typhoon spatiotemporal evolution model is established. The model can capture the key characteristics of typhoons, such as the typhoon radius, central wind speed, and circulation wind speed at different locations, so as to achieve real-time simulation of the typhoon movement path and wind field distribution. Through this coordinate system, the lines and equipment in the distribution network, such as wind turbines, are given specific coordinate positions. Combined with the typhoon spatiotemporal evolution model, it can be determined whether the distribution network lines and equipment will be affected by typhoons at different time stages, and the actual wind speed they may suffer. By analyzing the strength and load effects of the line elements, the time-varying failure rate of the distribution network lines can be calculated. Similarly, by studying the wind turbine structure and the wind speed model under severe convective weather, the time-varying failure rate of the wind turbine can be calculated. Furthermore, the lines and wind turbine equipment of the distribution network in the region are equivalent to different nodes, and according to the types of different nodes, their distribution network line failure rates or wind turbine failure rates at different times of the typhoon are calculated to form a time-varying failure rate sequence as the input of the graph attention method. The powerful feature learning mechanism and the ability to capture complex nonlinear relationships of the graph attention method are used to construct a distribution network failure model to predict the failure rate of the distribution network. In model training, the pinball loss function is selected as the loss function of the model, and the Adam optimizer is used to train the model. The goal is to minimize the loss function, so as to obtain an accurate prediction of the distribution network failure rate. At the same time, the uncertainty of wind power output is also a factor that cannot be ignored in typhoon weather. In order to deal with this uncertainty, the distributed robust optimization method is introduced to construct a spherical fuzzy set based on the historical wind power output data set, and combined with the predicted distribution network failure rate, a trading model considering uncertain distributed resources is established. The allocation and trading of power resources can be optimized under extreme weather conditions to ensure the stable operation of the power grid.

[0068] The specific steps of S1 are:

[0069] The distribution area is divided into grids at the geographical level. After the grids are evenly divided, the number of typhoon durations is set as T according to the meteorological data of GIS, and the model of the spatiotemporal evolution of typhoon disasters is established;

[0070] S1.1 Calculate the coordinates of the typhoon center point in the grid at time t as follows:

[0071]

[0072] Where (x 0 ,y 0 ) is the position of the typhoon at time t = 0, (x T ,y T ) is the position of the typhoon at time t = T;

[0073] S1.2 Calculate the radius R of the typhoon level 10 wind circle at time t 10,t for:

[0074]

[0075] Where R 10,0 , R 10,T They are the radii of the typhoon level 10 wind circle at t=0 and t=T respectively;

[0076] S1.3 Calculate the maximum wind speed V at the typhoon center at time t o,max,t for:

[0077]

[0078] Where V o,max,0 、V o,max,T are the maximum wind speed at the typhoon center at t=0 and t=T respectively;

[0079] S1.4 Calculate the maximum wind speed V max

[0080] Maximum wind speed V in the wind farm max It is described by extreme value type III (Weibull) distribution, and its probability density function is shown below. The fitting error is about 7.7%.

[0081]

[0082] Where: a is the scale parameter; γ is the shape parameter; b is the location parameter. The fitting values ​​are a=11.8174, b=23.1902, γ=4.8878,

[0083] S1.5 Calculate the maximum circulation wind speed V r,max for:

[0084]

[0085] S1.6 Calculation of circulation wind speed V r,t for:

[0086]

[0087] The typhoon spatiotemporal evolution model simulates the state of a typhoon when it passes through. Since a coordinate system is constructed for the region, the typhoon wind circle radius and maximum wind speed can be used to understand whether the equipment or lines at different locations are affected by the typhoon and the actual wind speed. In typhoon weather, wind speed is the main factor causing failures. The typhoon spatiotemporal evolution model can be used to obtain the actual wind speed at the location of the lines and equipment, and then the time-varying failure rate of the distribution network lines and the time-varying failure rate of the wind turbines can be calculated.

[0088] The specific steps of S2 are:

[0089] The distribution network lines and wind turbine equipment in the region are equivalent to nodes in the graph attention method;

[0090] Relying on the model of typhoon spatiotemporal evolution, the wind force of the node in different time periods is obtained from the node's location coordinates, and the corresponding distribution line time-varying failure rate or wind turbine time-varying failure rate is calculated according to different node types.

[0091] S2.1 Calculation of the time-varying fault rate P of the distribution network line L ;

[0092] The time-varying fault rate of distribution network towers in typhoon weather is basically the same as that of distribution network conductors, but its fault rate is much lower than that of distribution network conductors, with a difference of 10 -3 The order of magnitude is such that the impact of tower collapse on the fault scenario of the distribution network can be ignored, and thus the line failure rate of the distribution network is equivalent to the conductor failure rate.

[0093] The failure rate of the conductor is mainly determined by the relationship between the inherent strength of the component and the load it bears, ignoring the wind load along the conductor direction. w The vertical wind load on the conductor per unit length is for:

[0094]

[0095] Where: D is the outer diameter of the conductor; α is the wind pressure unevenness coefficient, which is set to 0.61 in this paper; μ sc is the wind load shape coefficient, when D<17mm, μ sc Take =1.2, when D≥17mm, take μ sc =1.1;μ Z is the wind pressure height variation coefficient, μZ =1.0; L H is the horizontal spacing of the towers; θ is the angle between the wind direction and the conductor direction.

[0096] Without considering the pressure model caused by heavy rain brought by typhoon, the comprehensive load per unit length of wire in typhoon weather is:

[0097]

[0098] In the formula, is the vertical load component of the conductor per unit length, that is, the load caused by its own gravity. From this, the stress σ on the conductor cross section can be obtained g , stress σ g Proportional to

[0099] The probability distribution of the ultimate strength of the material should adopt the normal distribution, and the failure rate of the distribution line is:

[0100]

[0101] The highest suspension point of the overhead wire is prone to wire breakage failure, and the stress on the wire cross section σ g Proportional to the sum of the wind load and gravity load on the conductor. Where: μ 1 , δ 1 are the mean and standard deviation of the tensile strength of the wire respectively; 1 Indicates the breaking stress of the conductor.

[0102] S2.2 Calculation of time-varying failure rate of wind turbines

[0103]

[0104] Where: It indicates the probability of accidental failure of the fan under normal operating conditions, which can be obtained through the historical data of the fan operation failure rate. Under normal weather conditions, the probability of the fan failure state is 7.3%; v r 、v cutin 、v cutoff are the rated wind speed, cut-in wind speed and cut-out wind speed of the fan respectively; v w is the real-time wind speed at the wind farm’s location. Figure 2 shown.

[0105] The specific steps of S3 are:

[0106] S3.1 The first stage graph attention method predicts the failure rate of the distribution network:

[0107] The distribution network lines and wind turbine equipment in the region are equivalent to the nodes in the graph attention method. Since the coordinate system is established and each node has corresponding coordinates, the actual wind speed of the node in each time period can be equivalently obtained by referring to the typhoon spatiotemporal evolution model. The time-varying failure rate of the distribution network line or the time-varying failure rate of the wind turbine at the node in different time periods is calculated according to the corresponding failure rate formula based on the node type.

[0108] The time-varying failure rate sequence corresponding to the calculated nodes is used as the input of the graph attention method:

[0109]

[0110] The combination of n nodes is:

[0111] X=[X 1 ,X 2 ,...,X n ]

[0112] A node v in the distribution network i , there will be many neighboring nodes v around j In order to better distribute the weights, the correlations calculated for all adjacent nodes are normalized by softmax to obtain the attention coefficient a ij :

[0113]

[0114] Where: L represents the activation function LeakyReLU; α represents the function for calculating the correlation between two nodes; W represents the weight parameter matrix of the node from the input feature dimension to the output feature dimension.

[0115] After obtaining the attention coefficient, we can get the node V according to the idea of ​​weighted summation. i The output feature vector of is:

[0116]

[0117] Where: Y i Represents the node v of this layer i The new feature vector of ; σ represents the activation function, usually the eLU function is used. The mathematical expression of the eLU function is:

[0118]

[0119] Here, α is a hyperparameter and is usually set to 1.

[0120] The multi-head attention mechanism obtains more comprehensive information and enhances the stability of the model by independently calculating M groups of attention. The splicing operation is used in the middle layer of GAT to improve the expression ability of the attention layer, and the last layer adopts the average operation.

[0121]

[0122] Where: M is the number of attention heads; || represents the concatenation operation; and W m are the weight coefficient and learning parameter of the mth group of attention mechanism respectively.

[0123] S3.2 The second stage graph attention method obtains the probability distribution of failure rate:

[0124] In the second stage, the graph attention method selects the pinball loss function as the loss function of the model, and uses the Adam optimizer to train the model so that the loss function reaches the optimal solution and finally obtains the predicted failure rate.

[0125]

[0126] Among them, y * is the actual value, y q is the model prediction value, q is the quantile cumulative probability corresponding to the pinball loss function, which corresponds to the range of 0.5% to 99.5%.

[0127] The specific steps of S4 are:

[0128] S4.1 will be based on the sample collection of historical data The empirical distribution P of the wind power prediction error obtained K As an estimate of the true distribution P;

[0129] S4.2 Calculation of empirical distribution P K The Wasserstein distance between these two discrete distributions and the true distribution P:

[0130]

[0131] Where: Uncertain parameters Obey P K and P; Π is P K The joint distribution of and P.

[0132] S4.3 Constructing the empirical distribution P K The spherical fuzzy set D with the center as the center and the Wasserstein probability distance as the radius 0 ;

[0133] Fuzzy uncertain set D 0 Defined as:

[0134]

[0135] In the formula represents the total probability distribution over Ξ. Note that D 0 is an empirical distribution P K The Wasserstein ball with ε(K) as the radius contains all possible probability distributions with a certain degree of confidence. ε(K) is the probability distribution of the fuzzy uncertain set D. 0 Have an important impact and should meet ε(K) is:

[0136]

[0137] Where: β is the confidence level; D is the coefficient; it can be obtained by solving the following optimization problem:

[0138]

[0139] In the formula is the sample mean.

[0140] The specific steps of S5 are:

[0141] S5.1 Construct the objective function of the distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons:

[0142]

[0143] C grid (t) = C buy (t)-C sell (t)

[0144] C buy (t) = c buy (t)P buy (t)

[0145] C sell (t) = c sell (t)P sell (t)

[0146]

[0147] C wt (t) = o wt P wt (t)

[0148] C pv (t) = o pv P pv (t)

[0149] C bat (t) = o ch P ch (t)

[0150] Where: Cgrid The cost of the load aggregator interacting with the main network, including the cost of purchasing electricity C buy and electricity sales revenue C sell , where c buy 、c sell are the electricity purchase and sale prices, P buy , P sell They are respectively the power of electricity purchased and the power of electricity sold; C MT is the gas turbine power generation cost, where a i 、b i 、c i is the fuel coefficient; C wt , C pv , C bat are the operation and maintenance costs of wind turbines, photovoltaic units and battery energy storage modules, among which o wt , o pv , o ch are the unit power operation and maintenance costs of wind turbines, photovoltaic units and batteries, P wt , P pv , P ch (t) are the power generation of wind turbine, photovoltaic unit and battery respectively;

[0151] S5.2 sets the constraints of the distributed resource uncertainty trading strategy considering the temporal and spatial evolution of typhoons. The constraints include power flow constraints, distributed flooding opportunity constraints, gas turbine output power and ramp constraints, power main grid output power constraints, energy storage equipment constraints, and power balance constraints:

[0152] AC power flow equation constraints:

[0153]

[0154] Line power constraints:

[0155]

[0156] Operation constraints:

[0157]

[0158] Distributed stick chance constraints:

[0159]

[0160] In the above formula: B and L represent the busbar set and line set respectively; G i,j +i·B i,j is the element of the system admittance matrix; in They represent the output of the gas turbine, wind farm, photovoltaic power station, and energy storage equipment of node i in time period t. When there is no corresponding unit at node i, its value is 0; represents the active load of node i; in It represents the output of reactive power of node i in time period t. When node i has no reactive power, its value is 0. represents the reactive load of node i; represents the power of line ij in time period t. i,j,t Indicates the fault status of the line at time t, usually a binary variable, indicating whether the line is faulty (1 indicates fault, 0 indicates normal). represents the active power of the wind farm at time t, Indicates the lower and upper limits of the active power of the wind farm at time t. P Is the wind farm in the fuzzy set D 0 The expectation of the next output, ∈, is used to represent the allowed risk probability. In the distributed robust chance constraint, it defines the upper limit of the probability that the constraint condition is not satisfied.

[0161] Gas turbine output power and ramp constraints:

[0162]

[0163] P MT (t)-P MT (t-1)|≤r MT

[0164] Where: are the minimum and maximum power generation of the gas turbine respectively; r MT is the climbing coefficient.

[0165] Output power constraints of the power grid:

[0166]

[0167] Where: They are the upper limits of power purchase and sales by the main power grid, to prevent excessive fluctuations from bringing new safety hazards to the operation of the power grid.

[0168] Energy storage equipment constraints:

[0169]

[0170] SOC min (t)≤SOC(t)≤SOC max (t)

[0171] Where: They are the upper and lower limits of the charging and discharging power of the energy storage battery, SOCmin , SOC max They are the upper and lower limits of battery capacity, respectively, to prevent the energy storage battery from being discharged too deeply and affecting its service life.

[0172] Power balance constraints

[0173] P grid (t)+P wt (t)(1-E wt (t))+P MT (t)(1-E line (t))+P bat (t)=Load(t)

[0174] Among them, E wt 、E line are the time-varying failure rate of wind turbines and the time-varying failure rate of distribution network lines in the actual failure rate prediction value under typhoon scenarios. Line failure will cause the gas turbine output to be unable to be sent to the power grid, affecting the operating environment of the gas turbine and the stability of the power grid, thereby indirectly affecting the operation and scheduling of the gas turbine. grid (t) is the main power grid power, P bat (t) represents the energy storage battery power, and Load(t) represents the load power.

[0175] S5.3 solves the constructed distributed resource uncertain trading strategy model considering the temporal and spatial evolution of typhoons, and obtains the output results of distributed resources after optimization and clearing, as shown in Figure 3 . Combined Figure 4 It can be seen that the impact of extreme weather events such as typhoons on distributed resource trading strategies is mainly reflected in the increase of uncertainty factors, which leads to an increase in prediction errors, thus having a negative impact on the economy of the system. This embodiment takes into account that typhoon weather will increase the prediction error of distributed resources, and processes the uncertainty of wind power output, and the actual value obtained is usually smaller than the predicted value. In this case, in order to make up for the error, the gas turbine needs to increase its output, which in turn leads to an increase in the transaction cost of the system. Therefore, when incorporating uncertainty factors into regional power market transactions, additional costs must be reserved to deal with potential emergencies. This embodiment minimizes the economic dispatch cost of the distribution network during typhoons by solving the distributed resource uncertainty trading strategy model that takes into account the spatiotemporal evolution of typhoons, optimizes the allocation and trading of power resources, and ensures the stable operation of the power grid.

[0176] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons, characterized in that: The strategy includes the following steps: S1. Establish a model for the temporal and spatial evolution of typhoon disasters; S2. Calculate the time-varying failure rate of nodes based on the model of the spatiotemporal evolution of typhoon disasters, wherein the time-varying failure rate of nodes includes the time-varying failure rate of distribution network lines and the time-varying failure rate of wind turbines. S3. Train the graph attention network based on the time-varying failure rate of the node to obtain the distribution network failure rate prediction model of the node based on the graph attention, and obtain the actual failure rate prediction value based on the distribution network failure rate prediction model. During the training process, the pinball loss function is used. S4, obtain the historical wind power output data set and construct the uncertainty fuzzy set based on Wasserstein distance; S5. Based on the uncertainty fuzzy sets and the actual failure rate prediction value, a distributed resource uncertain trading strategy model considering the spatiotemporal evolution of typhoons is constructed, and the trading strategy is obtained by solving the model.

2. According to claim 1, a distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons is characterized in that: The distribution network line time-varying fault rate P L for: Where: μ1 and δ1 are the mean and standard deviation of the tensile strength of the conductor respectively; σ1 represents the breaking stress of the conductor; σ g is the stress on the conductor cross section.

3. According to claim 2, a distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons is characterized in that: The time-varying failure rate of the fan for: in, It indicates the probability of accidental failure of the fan under normal operating conditions, which is obtained through the historical data of the fan operation failure rate; v r 、v cutin 、v cutoff are the rated wind speed, cut-in wind speed and cut-out wind speed of the fan respectively; v w It is the real-time wind speed at the geographical location of the wind farm.

4. According to claim 1, a distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons is characterized in that: The specific steps of training the graph attention network based on the time-varying failure rate of nodes are: The time-varying failure rate sequence of the distribution network line or the time-varying failure rate sequence of the wind turbine corresponding to each of the n nodes is used as the input of the graph attention network; Calculate the correlation between nodes and normalize them to get the attention coefficient; Get the output feature vector of the node based on the attention coefficient; The middle layer of the graph attention network concatenates the output feature vectors of the nodes, and the last layer of the graph attention network averages the output feature vectors of the nodes. The output of the last layer passes through the prediction head to obtain the model prediction value; The pinball loss function is calculated based on the model prediction value, and the graph attention network is trained based on the pinball loss function.

5. According to claim 4, a distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons is characterized in that: The pinball loss function is: Among them, y * is the actual value, y q is the model prediction value, and q is the cumulative probability of the quantile corresponding to the pinball loss function.

6. According to claim 1, a distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons is characterized in that: The specific steps of S4 are: Obtain the historical wind power output data set and convert the data set into the empirical distribution P of the wind and solar forecast error K As an estimate of the true distribution P; Calculate the empirical distribution P K The Wasserstein distance from the true distribution P; Construct the empirical distribution P K is the spherical fuzzy set D0 with ε(K) as the center and ε(K) as the radius of the Wasserstein ball.

7. The distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons according to claim 6 is characterized in that: The spherical fuzzy set D0 is: in, represents the total probability distribution on Ξ, W(P K ,P) represents the empirical distribution P K The Wasserstein distance between the actual distribution P and ε(K) is the empirical distribution P. K The radius of the Wasserstein ball of the fuzzy set D0 centered at ε(K) satisfies Among them, β is the confidence level; K represents the sample size; and D is the coefficient.

8. The distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons according to claim 1 is characterized in that: The objective function of the distributed resource uncertain transaction strategy model is: C grid (t)=C buy (t)-C sell (t) C buy (t)=c buy (t)P buy (t) C sell (t)=c sell (t)P sell (t) C wt (t)=o wt P wt (t) C pv (t)=o pv P pv (t) C bat (t)=o ch P ch (t) Among them, f1 represents the objective function, C grid The cost of the load aggregator interacting with the main network, including the cost of purchasing electricity C buy and electricity sales revenue C sell , where c buy 、c sell are the electricity purchase and sale prices, P buy , P sell They are respectively the power of electricity purchased and the power of electricity sold; C MT is the cost of gas turbine power generation, where a i , b i 、c i is the fuel coefficient; P MT represents the power generated by the gas turbine, C wt , C pv , C bat are the operation and maintenance costs of wind turbines, photovoltaic units and battery energy storage modules, among which o wt , o pv , o ch are the unit power operation and maintenance costs of wind turbines, photovoltaic units and batteries, P wt , P pv , P ch They are the power generation capacity of wind turbine, photovoltaic unit and battery respectively.

9. The distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons according to claim 8 is characterized in that: The constraints of the distributed resource uncertain trading strategy model include: AC power flow equation constraints, line power constraints, operation constraints, distributed blue-power opportunity constraints, gas turbine output power and ramp constraints, power main grid output power constraints, energy storage equipment constraints and power balance constraints.

10. The distributed resource uncertainty trading strategy considering the spatiotemporal evolution of typhoons according to claim 9, characterized in that: The distribution constraint is: Among them, the spherical fuzzy set D0, L represents the line set, z i,j,t represents the fault status of the line at time t, z i,j,t Indicates the fault status of the line at time t, usually a binary variable, indicating whether the line is faulty (1 indicates fault, 0 indicates normal). represents the active power of the wind farm at time t, Indicates the lower and upper limits of the active power of the wind farm at time t. P is the expected output of the wind farm under the fuzzy set D0, ∈ is used to represent the allowed risk probability. In the distributed robust chance constraint, it defines the upper limit of the probability that the constraint condition is not satisfied, ∈ is the violation probability; The power balance constraint is: P grid (t)+P wt (t)(1-E wt (t))+P MT (t)(1-E line (t))+P bat (t)=Load(t) Among them, E wt 、E line are the time-varying failure rate of wind turbines and the time-varying failure rate of distribution network lines in the actual failure rate prediction value under typhoon scenarios. Line failure will cause the gas turbine output to be unable to be sent to the power grid, which will affect the operating environment of the gas turbine and the stability of the power grid, thereby indirectly affecting the operation and scheduling of the gas turbine. grid (t) is the main power grid power, P bat (t) represents the energy storage battery power, and Load(t) represents the load power.

Images