CN112001490A - Method and system for determining confidence capacity of grid-connected photovoltaic system - Google Patents

Abstract

The invention discloses a method and a system for determining confidence capacity of a grid-connected photovoltaic system, wherein the method comprises the following steps: obtaining photovoltaic permeability, time step and photovoltaic load related similarity indexes; substituting the photovoltaic permeability, the time step and the photovoltaic load related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system; the empirical model is obtained by training the mapping relation between the confidence capacity and the photovoltaic permeability, the time step and the photovoltaic load related similarity index by using an artificial neural network. The present invention can estimate the confidence capacity for any given photovoltaic permeability, time step and photovoltaic load related similarity index without the need to use complex, time-consuming inverse monte carlo SMC calculations.

Description

A method and system for determining the confidence capacity of a grid-connected photovoltaic system

technical field

The invention relates to the technical field of power generation, in particular to a method and system for determining the confidence capacity of a grid-connected photovoltaic system.

Background technique

In recent years, in order to meet the increasing energy consumption and achieve a sustainable environment, large-scale grid-connected photovoltaic power plants have risen rapidly. With the increase of photovoltaic penetration, photovoltaic power plants not only contribute power value to the power system or distribution network, but also contribute to capacity value, so the definition of confidence capacity is proposed. At present, photovoltaic power generation has been widely used, and confidence capacity assessment is an important issue in photovoltaic power generation planning and scheduling.

There are currently reliability-based methods and approximations to assess PV confidence capacity. The method of confidence capacity evaluation based on reliability indicators such as power shortage probability, power shortage expectation value and power shortage expectation value is to evaluate the confidence capacity of photovoltaic power plants from the perspective of the reliability contribution of photovoltaic power plants to the power system. The reliability-based methods mainly include the effective carrying capacity method, the equivalent conventional power method and the equivalent enterprise power method. The effective carrying capacity method uses the effective load carrying capability (ELCC) to measure the confidence capacity of the photovoltaic system. The ELCC characterizes the capacity that the system load can increase while maintaining the reliability before and after the intermittent power generation is connected to the system. This indicator can intuitively represent the new load capacity that can be added after the system adds a generator set. In practice, the confidence capacity is usually measured by the capacity that the connected intermittent energy can reduce the planned conventional power generation. In the process of measuring the confidence capacity, the process of calculating the reliability using the reverse Monte Carlo SMC method is complex and time-consuming, resulting in Computational efficiency is low.

SUMMARY OF THE INVENTION

In order to solve the above deficiencies in the prior art, the present invention provides a method for determining the confidence capacity of a grid-connected photovoltaic system, including:

Obtain PV penetration rate, time step and PV load related similarity index;

Bringing the photovoltaic penetration rate, time step and photovoltaic load-related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

Wherein, the empirical model is obtained by using artificial neural network to train the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index.

Preferably, the training of the empirical model includes:

Set multiple photovoltaic capacities;

For each photovoltaic capacity, simulate under different time steps to obtain the corresponding photovoltaic output power distribution;

A sample set is formed based on all the time steps corresponding to each photovoltaic capacity and the photovoltaic output power distribution;

Build artificial neural networks based on different PV penetration rates, time steps and PV load-related similarity values;

The artificial neural network is trained based on the sample set, and the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index is obtained.

Preferably, the mapping relationship between the confidence capacity and the photovoltaic penetration rate, the time step and the photovoltaic load-related similarity index in the empirical model is shown in the following formula:

κ _CC =g(r,Δt,η)

In the formula: κ _CC is the confidence capacity of the photovoltaic system; r is the photovoltaic penetration rate; Δt is the time step; η is the photovoltaic-load correlation similarity index in time order.

Preferably, the photovoltaic-load related similarity index η is shown in the following formula:

η=v× _γC

In the formula: v is the ramp rate index of photovoltaic output power according to time change, γ _C is the correlation between photovoltaic output power and load distribution.

Preferably, the time-varying ramp rate index v of the photovoltaic output power is shown in the following formula:

为负荷时间序列；

为在时间步长Δt中光伏输出和负荷的绝对最大波动；

表示光伏输出功率时间序列的归一化斜坡率；

表示第i+1个光伏输出单元在时间步长Δt时对应的光伏容量；

表示第i个光伏输出单元在时间步长Δt时对应的光伏容量；

表示T时间内的归一化负荷斜坡率；L_(i+1)Δt表示第i+1个光伏输出单元在时间步长Δt时光伏输出和负荷的波动；L_iΔt表示第i个光伏输出单元在时间步长Δt时光伏输出和负荷的波动。In the formula: n represents the number of photovoltaic output units;

is the load time series;

is the absolute maximum fluctuation of PV output and load in time step Δt;

Represents the normalized ramp rate of the photovoltaic output power time series;

Represents the photovoltaic capacity corresponding to the i+1th photovoltaic output unit at the time step Δt;

represents the photovoltaic capacity corresponding to the i-th photovoltaic output unit at the time step Δt;

Represents the normalized load ramp rate within T time; L _(i+1)Δt represents the fluctuation of PV output and load of the i+1th PV output unit at time step Δt; L _iΔt represents the i-th PV output unit Fluctuations in PV output and load at time step Δt.

Preferably, the artificial neural network is a back-propagation neural network.

Based on the same inventive concept, the present invention also provides a system for determining the confidence capacity of a grid-connected photovoltaic system, including:

The acquisition module is used to obtain the PV penetration rate, time step and PV load related similarity index;

a determination module, used for bringing the photovoltaic penetration rate, time step and photovoltaic load related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

Wherein, the empirical model is obtained by using artificial neural network to train the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index.

Preferably, the system further includes a training module; the training module is specifically used for:

Set multiple photovoltaic capacities;

For each photovoltaic capacity, simulate under different time steps to obtain the corresponding photovoltaic output power distribution;

A sample set is formed based on all the time steps corresponding to each photovoltaic capacity and the photovoltaic output power distribution;

Build artificial neural networks based on different PV penetration rates, time steps and PV load-related similarity values;

The artificial neural network is trained based on the sample set, and the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index is obtained.

Preferably, the mapping relationship between the confidence capacity and the photovoltaic penetration rate, the time step and the photovoltaic load-related similarity index in the empirical model is shown in the following formula:

κ _CC =g(r,Δt,η)

In the formula: κ _CC is the confidence capacity of the photovoltaic system; r is the photovoltaic penetration rate; Δt is the time step; η is the photovoltaic load correlation similarity index in time order.

Preferably, the photovoltaic load related similarity index η is shown in the following formula:

η=v× _γC

In the formula: v is the ramp rate index of photovoltaic output power according to time change, γ _C is the correlation between photovoltaic output power and load distribution.

Compared with the closest prior art, the technical solution provided by the present invention has the following beneficial effects:

In the technical solution provided by the present invention, after the photovoltaic permeability, time step and photovoltaic load related similarity index are obtained; the photovoltaic permeability, time step and photovoltaic load related similarity index are brought into a pre-trained empirical model to determine the photovoltaic system The confidence capacity of ; the empirical model in this technical solution is obtained by using artificial neural network to train the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index. The present invention can estimate the confidence capacity for any given PV penetration, time step, and PV load-related similarity index without the need to use complex, time-consuming inverse Monte Carlo SMC calculations.

Description of drawings

1 is a flowchart of a method for determining the confidence capacity of a grid-connected photovoltaic system provided by the present invention;

Fig. 2 is the secant method schematic diagram of ELCC evaluation provided by the present invention;

3 is a schematic diagram of an artificial neural network architecture for photovoltaic confidence capacity evaluation provided by the present invention;

Fig. 4 is the schematic diagram of ELCC iterative simulation in the embodiment of the present invention;

5 is a schematic diagram of a fitting curve between photovoltaic confidence capacity and η index in an embodiment of the present invention;

6 is a schematic diagram of the time step of PV-load correlation-similarity index under different PV permeability in an embodiment of the present invention;

7 is a schematic diagram of the influence of photovoltaic permeability on confidence capacity in an embodiment of the present invention;

8 is a boxplot of photovoltaic confidence capacity at different time intervals of 100MW and 1000MW photovoltaic power plants in the embodiment of the present invention;

9 is a schematic diagram of the influence of time interval and photovoltaic permeability on photovoltaic confidence capacity in an embodiment of the present invention;

FIG. 10 is a schematic diagram of a photovoltaic confidence capacity evaluation result in an embodiment of the present invention.

Detailed ways

In order to better understand the present invention, the content of the present invention will be further described below with reference to the accompanying drawings and examples.

Embodiment 1: As shown in Figure 1, the present invention provides a method for determining the confidence capacity of a grid-connected photovoltaic system, including:

S1 obtains the PV penetration rate, time step and PV load related similarity index;

S2 brings the photovoltaic permeability, time step and photovoltaic load related similarity index into the pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

Wherein, the empirical model is obtained by using artificial neural network to train the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index.

The present invention first establishes a surface-of-array (POA) irradiance model. When the standard composition of irradiance at a certain time is known, the irradiance reaching the inclined photovoltaic module can be calculated according to the following formulas (1)-(5). The standard components of irradiance include: direct normal irradiance (DNI), global horizontal irradiance (GHI) and diffuse horizontal irradiance (DHI) of solar irradiance.

The POA irradiance E _POA is given by the following formula:

E _POA =E _b +E _g +E _d (1)

In the formula: E _b is the POA beam component, E _g is the POA ground reflection component, and _Ed is the POA diffusion component.

The POA beam component E _b can be obtained by the following formula:

E _b =DNI×cos(AOI) (2)

AOI=cos ^-1 [cos(θ _Z )cos(θ _T )+sin(θ _Z )sin(θ _T )×cos(θ _A -θ _α )] (3)

Where: AOI is the incident angle, θ _T is the inclination of the array, θ _a is the azimuth angle of the array, 0° north, 90° east, 180° south, and 270° west, in a clockwise direction. The sun azimuth angle θ _A and the zenith angle θ _Z can be derived from the Sun Position Algorithm (SPA) model described by Reda and Andreas.

The ground reflection component E _g can be obtained by the following formula:

Where: ρ is the albedo, which describes the reflectivity of the surface. The albedo represents the reflected GHI, with low reflectance on dark sides and high values on bright white sides. In the present invention, ρ=0.2 and ρ=0.8 are assumed for dry concrete and fresh snow on the ground, respectively.

The present invention uses the Sandia empirical model of sky scattered radiation to calculate the E _d component as follows:

where: θ _T is the inclination of the array, θ _Z is the zenith angle derived from the Sun Position Algorithm (SPA) model described by Reda and Andreas, and in equation (5), the first term is the isotropic sky scattering The radiation model, the second term is the empirical correction term, is used to account for the brightening effect around the sun and the horizon.

The existing photovoltaic power station has the characteristics of a low-pass filter. The present invention proposes a photovoltaic output model based on the low-pass filter. On the basis of frequency domain analysis, a first-order transfer function is proposed to describe the photovoltaic output power and irradiance. The nonlinear relationship between , as shown in the following formula:

In the formula: G(t) is the GHI time series, and P(t) is the simulated photovoltaic power output time series. Photovoltaic power station area S (ha) is the main factor affecting the stability of photovoltaic power generation fluctuations.

Converting the transfer function from analog to digital filter, the discrete transfer function can be written as:

z＝e^sf，f是实测辐照度时间序列的采样频率，P^*(W)是变压器标称功率，G^*(1000W/m²)代表参考辐照度。为了简化，光伏电站的组件被认为是绝对可靠的，而不考虑故障特性。本发明中的光伏功率密度，代表布满太阳能电池板的单位面积直流输出功率，地面固定倾斜度为65-dcW/m²，作为经验法则被采纳。因此，面积为1公顷的光伏电站的装机容量为P^*＝0.65MW。where:

z= ^esf , f is the sampling frequency of the measured irradiance time series, P ^* (W) is the nominal power of the transformer, and G ^* (1000W/m ² ) represents the reference irradiance. For simplicity, the components of a photovoltaic power plant are considered infallible, regardless of fault characteristics. The photovoltaic power density in the present invention represents the DC output power per unit area covered with solar panels, and the fixed slope of the ground is 65-dcW/m ² , which is adopted as a rule of thumb. Therefore, the installed capacity of a photovoltaic power plant with an area of 1 hectare is P ^* = 0.65MW.

The invention provides the reliability of the photovoltaic system based on reverse Monte Carlo (SMC) simulation, and uses the secant method to determine the effective load capacity (ELCC) of the photovoltaic system. Figure 2 shows the reliability curves of the basic power system, the power system with additional conventional generators, and the power system with photovoltaic generator sets, respectively.

The implementation steps of the secant method are as follows:

Step 1: Calculate the expected load shortfall value of the basic power system without photovoltaic generator sets using SMC simulation, record it as R ₀ and annual peak load L _pk0 and C _con of conventional generators, and use point S ( L _pk0 , R ₀ ) represents. Reliability R _A =R(C _con +C ^PV ,L _pk0 ) is calculated when adding C ^PV photovoltaic power generation units to the basic system, denoted by A(L _pk0 , R _A ). After the peak load of the year increases to L _pk0 +C ^PV , the reliability of the photovoltaic system with C ^PV can be expressed as R _B =R(C _CON +C ^PV ,L _pk0 +C ^PV ), and B(L _pk0 +C ^PV ,R _B ) said.

Step 2: Calculate the intersection S ₁ of the straight line segment AB and f(x)=R ₀ , and determine the corresponding annual peak load L ₁ . Estimate the new reliability level R _C =R(C _CON +C ^PV ,L ₁ ) by SMC simulation, if |R _C -R ₀ |>ε, continue to calculate the intersection of line segment BC and f(x)=R ₀ S ₂ , replace A with C. This process is run iteratively until |R _C - R ₀ | > ε, where ε = 0.001 is a desired error threshold.

Step 3: Determine the effective load capacity ELCC of the PV system. When the convergence condition is satisfied, assuming that the intersection point is D(L _D , R ₀ ), the calculated peak load can be expressed as L _D =L _pk0 +ΔL. This means that the installed C ^PV photovoltaic system can take the additional ΔL load and maintain the specified reliability level. The capacity value is determined by the difference between the annual peak load of the basic system L _pk0 and the annual peak load LD of the photovoltaic power generation system _. The confidence capacity can be determined by (8). Obviously, the additional capacity value ΔL of the photovoltaic power station belongs to the range of [0, C ^PV ] .

In the above steps, the connected intermittent energy can reduce the planned conventional power generation capacity to measure the confidence capacity. At this time, the ELCC is not directly used to measure the confidence capacity of the newly added intermittent energy.

Through the above three steps, it can be seen that according to the reverse Monte Carlo (SMC) simulation to calculate the reliability of the photovoltaic system, it is necessary to use the secant method to determine the effective load capacity (ELCC) of the photovoltaic system, which includes: firstly, it is necessary to determine the basic power system, The reliability curve of the power system with conventional generators and the power system with photovoltaic generator set is added, and then the intersection point of the curve is determined, and the intersection point S ₁ of the straight line segment AB and f(x)=R ₀ is calculated, and the corresponding annual peak value is determined load L ₁ . Estimate the new reliability level R _C =R(C _CON +C ^PV ,L ₁ ) by SMC simulation, if |R _C -R ₀ |>ε, continue to calculate the intersection of line segment BC and f(x)=R ₀ S ₂ , replace A with C. This process is repeated until |R _C -R ₀ |>ε. The SMC calculation is not only complicated in the calculation process, but also it is necessary to repeatedly estimate the new reliability level through the SMC simulation when calculating the confidence capacity of the photovoltaic system. R _C =R( _{C CON} +C ^PV ,L ₁ ), if | ₀ |>ε, then continue to calculate the intersection S ₂ of the line segment BC and f(x)=R ₀ , and replace A with C, which takes a lot of time.

The present invention proposes a new metric that simultaneously describes the variability and time dependence of photovoltaic output and load distribution, in which a new time-varying ramp rate index v of photovoltaic output is introduced:

表示负荷时间序列，

表示PV输出和负荷的绝对最大波动，在时间步长Δt中。分子项表示PV功率时间序列的归一化斜坡率，分母项表示T时间内的归一化负荷斜坡率。In the formula: n represents the number of PV output units,

represents the load time series,

Represents the absolute maximum fluctuation of PV output and load, in time step Δt. The numerator term represents the normalized ramp rate of the PV power time series, and the denominator term represents the normalized load ramp rate over time T.

A chronological PV-load-related similarity index η is defined, which describes the main factors affecting the ELCC evaluation of photovoltaic systems, and is defined as follows:

η=v× _γC (10)

η is the PV-load correlation similarity index in chronological order, v is the time-varying ramp rate index, that is, the rate of change of PV output; γ _C is the correlation between PV output and load distribution.

In practice, the confidence capacity κ _CC of a photovoltaic system is mainly determined by the PV permeability r, the time step Δt, the rate of change of the photovoltaic output v, and the correlation between the photovoltaic output and the load distribution γ _C. The specific performance is as follows:

κ _CC =f( _r ,Δt,v,γc ) (11)

For different PV penetration rates, time scales and correlations, a back-propagation (BP) neural network is established to estimate the confidence capacity of the photovoltaic system. The empirical model of the artificial neural network is used to calculate the photovoltaic confidence capacity. The empirical model is as follows:

κ _CC =g(r,Δt,η) (12)

where r is the PV permeability, Δt is the time step and the PV-load correlation-similarity index η.

The empirical model provided by the present invention is constructed based on artificial neural network, and its construction steps are:

(1) Select different PV capacities and conduct extensive simulations at various simulation time intervals to obtain artificial neural network (ANN) sample sets. at PV capacity vectors (10, 50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000) MW and at various simulation time intervals (2, 3, 4, 5, 7, 10, 15, 20, 30, 35, 40, 45, 50, 60 minutes) extensive simulations were carried out, 30 scenarios were obtained by random permutation of daily PV output when simulating at the specified PV capacity and at the specified time step, Randomly transform the 364-day photovoltaic output time series on the daily time scale to obtain different photovoltaic output distributions, and obtain a sample set of 5040 units through simulation;

(2) The back-propagation (BP) neural network established for different photovoltaic penetration rates, time scales and photovoltaic load-related similar values is used to estimate the confidence capacity of photovoltaic systems. train;

(3) After the training is completed, all connection weights of the network are recorded, an empirical model is established, and then the PV confidence capacity is simulated and evaluated with the trained network.

The relationship between confidence capacity and different PV penetration rates, time steps and PV load-related similarity values in the empirical model is shown as follows:

κ _CC =g(r,Δt,η) (13)

where r is the PV permeability; Δt is the time step; η is the PV load-related similarity index.

In this embodiment, there is a photovoltaic penetration rate=photovoltaic capacity/total capacity; the photovoltaic load-related similarity index needs to be obtained according to the photovoltaic output power distribution and load distribution.

The present invention considers the intermittency of photovoltaic power generation, and proposes a new measurement method to describe the time correlation between photovoltaic output and load distribution, that is, the ramp rate index v that changes according to time, which effectively verifies photovoltaic penetration, simulation time The influence of factors such as granularity, correlation between PV and load time series on PV system confidence capacity assessment.

The confidence capacity assessment method proposed by the invention can be applied to confidence capacity assessment under arbitrary conditions, verifies the law of decreasing marginal capacity value, and obtains the optimal time scale of confidence capacity assessment by using the established model.

This embodiment uses the IEEE RTS-79 system as the basic power system, and has 32 conventional generators with a capacity of 3405MW and a peak load of 2850MW. The annual load data, capacity and availability of conventional generators are obtained from the IEEE RTS-79 system, and the average 3s irradiance is obtained by moving time windows over a given time interval. In the reliability calculation, the linear interpolation method is used to determine the load value. Given the load and PV time series, only the SMC simulation is used to sample the output state of conventional generators when performing reliability calculations.

Utilize the empirical model based on artificial neural network provided by the present invention to evaluate confidence capacity, and its steps are as follows:

PV penetration is a key factor in ELCC estimation, in general, PV plants have high confidence capacity with low PV penetration, and low confidence capacity values with high PV penetration.

PV penetration is defined as follows:

Due to the intermittent nature of solar irradiance, the optimal PV confidence capacity evaluation interval should be selected.

In addition, in order to reflect the influence of PV and load changes on the ELCC evaluation, a new time-varying ramp rate index v is proposed:

表示负荷时间序列，

表示PV输出和负荷的绝对最大波动，在时间步长Δt中。分子项表示PV功率时间序列的归一化斜坡率，分母项表示T时间内的归一化负荷斜坡率。In the formula: n represents the number of PV output units,

represents the load time series,

Represents the absolute maximum fluctuation of PV output and load, in time step Δt. The numerator term represents the normalized ramp rate of the PV power time series, and the denominator term represents the normalized load ramp rate over time T.

和V₂＝{L_Δt,L_2Δt,...,L_nΔt}分别是PV输出和负荷需求的时间序列。本发明引入Spearman秩相关系数ρ_S来描述V₁和V₂之间的相关性。The correlation between PV and load distribution is also an important factor affecting the credit evaluation of photovoltaic power generation capacity.

and V ₂ ={L _Δt , L _2Δt , . . . , L _nΔt } are the time series of PV output and load demand, respectively. The present invention introduces the Spearman rank correlation coefficient ρ _S to describe the correlation between V ₁ and V ₂ .

和L_iΔt在升序向量中的位置记录等级x_i和y_i。变量d_i＝x_i-y_i是等级x_i和y_i的差值，Spearman秩相关系数只是描述了两条曲线的变化趋势，而不考虑它们的平均距离。为了便于描述两个时间序列之间的平均距离，引入Frechet距离。Frechet距离是度量两条连续直线或曲线(V₁和V₂)之间的相似性的度量，公式(17)中V₁和V₂的离散Frechet距离δ_dF(V₁,V₂)提供了一个很好的连续度量的近似，可以用一个简单的算法实现：Sort the vectors V ₁ and V ₂ in ascending order and according to the variable

and L _iΔt record the rank x _i and y _i at the position in the ascending vector. The variable d _i = _xi -y _i is the difference between the levels x _i and y _i , and the Spearman rank correlation coefficient just describes the changing trend of the two curves without considering their average distance. In order to facilitate the description of the average distance between two time series, the Frechet distance is introduced. Frechet distance is a measure of the similarity between two consecutive straight lines or curves (V ₁ and V ₂ ), the discrete Frechet distance δ _dF (V ₁ ,V ₂ ) of V ₁ and V ₂ in Eq. (17) provides A good approximation of continuous metrics can be achieved with a simple algorithm:

δ _dF (V ₁ , V ₂ )=min{||D|||D is a coupling between V ₁ and V ₂ } (17)

_{The modulo} D between V1 and V2 follows the order of the _{midpoints of} V1 and V2, and the length of modulo D is the length of the longest link in D, i.e.

where the dist() function represents the Euclidean norm, the discrete Frechet distance is defined as the largest pointwise distance minimized across all parameterizations, and other pointwise distances on the curve have no effect on the Frechet distance. in order to account for all pointwise distances. The present invention chooses the mean Frechet distance definition. The average discrete Frechet distance δ _α (V ₁ , V ₂ ) can be defined as the sum of the discrete Frechet distances as follows:

和L'_I＝L_iΔt/max(V₁,V₂)分别是归一化的PV发电和负荷需求，在得到Spearman秩相关系数ρ_S和归一化平均离散Frechet距离δ_α后，定义了一个度量γ_C，它同时表示两条曲线之间的相关性和平均距离(或相似性)：in

and L' _I =L _iΔt /max(V ₁ , V ₂ ) are the normalized PV generation and load demands, respectively. After obtaining the Spearman rank correlation coefficient ρ _S and the normalized average discrete Frechet distance δ _α , we define A metric γ _C that represents both the correlation and the average distance (or similarity) between two curves:

If the value of ρ _S is high and the value of δ _α is small, that is, the PV distribution is basically consistent with the load distribution, and the two curves are close, then the PV output and the load demand are well matched, and in this case a higher PV confidence can be obtained capacity.

In practice, the ELCCκ _CC of a photovoltaic power station is mainly determined by the PV penetration rate r, the time step Δt, the rate of change of the photovoltaic output v, and the correlation between the photovoltaic output and the load distribution γ _C , which is specifically expressed as:

κ _CC =f( _r ,Δt,v,γc ) (21)

Considering the above factors, a chronological PV-load-related similarity index η is defined, describing the main factors affecting the ELCC evaluation of photovoltaic systems:

η=v× _γC (22)

Equation (21) can be rewritten as:

κ _CC =g(r,Δt,η) (23)

Here, η is a composite measure of time step, correlation, and mean discrete distance describing PV and load distributions.

Figure 3 shows the back-propagation (BP) neural network built for different PV permeability r, time step Δt and PV-load-related similarity index η to estimate the confidence capacity of photovoltaic systems. where w _ij represents the weight between nodes i and j, and the neuron’s bias is represented by θ _i (k) and θ _j (k). The present invention points out that the hidden layer neuron NHLN is only a possible choice, not an optimal choice, and a quantum (Levenberg-Marqudt, LM) back-propagation training algorithm is used to train the neural network. The hidden layer and output layer adopt the hyperbolic tangent Sigmoid transfer function "tansig" and the linear transfer function "purelin", respectively. Set the learning rate and max iterations to 0.05 and 3000.

To obtain the ANN sample set, at PV capacity vectors (10, 50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000) MW and at various simulation time intervals (2, 3, 4, 5, 7, 10, 15, 20, 30, 35, 40, 45, 50, 60 minutes) with extensive simulations. For each simulation (specified PV capacity and specified time step), 30 scenarios are obtained by random permutation of daily PV output. In order to obtain different PV output distributions, the 364-day PV output time series is randomly transformed on the daily time scale, A sample set of 5040 units is obtained by simulation. When the artificial neural network completes the training process, all the connection weights of the network are recorded, and the trained ANN has the ability of reasoning and generalization, and the empirical model proposed in equation (10) is established. With time step Δt and PV-load correlation-similarity index η, PV confidence capacity can be evaluated with the trained network without time-consuming SMC simulation.

Figure 4 shows the simulation process of confidence capacity evaluation for a 100MW photovoltaic system at a 1-h time step. In each iteration of the SMC simulation, the ELCC of the photovoltaic device was calculated and the ELCC curves are presented in Fig. 4. Assuming a photovoltaic power density of 65-dcW/m ² , the area of a 100 MW photovoltaic system is 153.8 hectares. The parameters estimated by ELCC of the 100MW photovoltaic power station are: r=0.0285, ν=0.1629, ρ _S =0.4683, δ _a =0.4523, γ _C =1.0353, η=0.1687. The present invention uses the Garver approximation method to calculate the photovoltaic confidence capacity, and compares it with the SMC simulation results. The calculation results are consistent with the calculation results of the proposed method.

To illustrate the relationship between the proposed PV load-related similarity index and PV confidence capacity, a multi-scenario simulation of the PV output curve and annual load distribution of the system in formula IEEE RTS-79 was performed. Each point in Figure 5 represents a scenario under a specific PV output time series, and in each scenario simulation, daily random permutation of PV output for each day was performed to obtain a specific PV output distribution. The fitting curve was obtained by using the linear fitting technique, and the fitting curve represented the positive correlation between the PV confidence capacity and the PV load correlation similarity index.

As shown in Fig. 6, the PV-load-related similarity η also varies with time interval, indicating that η varies with simulation time interval and PV permeability. It can be found that under different simulation time intervals, the PV-load correlation-similarity index increases with the increase of the simulation time interval; under the same simulation time step, the higher the PV permeability, the higher the PV-load correlation-similarity index. This is because the average distance between photovoltaic output and load demand is small due to the high photovoltaic penetration rate; the higher the photovoltaic penetration rate, the larger the power station area, and the more obvious the smoothing effect on the photovoltaic power generation output; the lower the photovoltaic penetration rate , the smaller the effect of the simulation time scale on the PV-load correlation-similarity index.

When the artificial neural network is trained, the generalization ability of ANN can be used to realize photovoltaic confidence capacity evaluation. Figure 7 illustrates the effect of PV permeability on the confidence capacity at 1-h time resolution. The research results are consistent, indicating that the PV confidence capacity will saturate with the increase of PV installed capacity, and a conclusion can be drawn that the marginal capacity value of the PV system decreases , in accordance with the law of diminishing marginal utility. The results in Table 1 show that the marginal capacity value of photovoltaic power generation decreases with the increase of photovoltaic installations.

Table 1 Decreasing marginal capacity values at 1-h time resolution of photovoltaic systems

Figure 8 shows the PV confidence capacity boxplots of 100MW and 1000MW PV plants at different time intervals. The simulation time step has a significant effect on the 100MW and 1000MW PV confidence capacity estimates. The results show that the choice of photovoltaic output time interval and load distribution has a great influence on the photovoltaic confidence capacity evaluation, and the optimal time step is also determined by the PV permeability.

From Figure 9, it can be found that the lower PV penetration rate (200MW PV capacity) leads to a larger confidence capacity variation range (30.1862%, 42.2492%) than that of the higher PV penetration rate (1000MW PV capacity) (8.0520%, 12.2028%). This is mainly due to the smoothing effect of large photovoltaic power plants.

Figure 10 is a trained ANN-based empirical model demonstrating the effect of time interval and PV penetration on PV capacity credit assessment. It can be found in Figure 10 that the results are somewhat different from Table 1, but still within the acceptable range. It can be found that the PV confidence capacity assessment depends on the choice of time interval, and as the time interval increases, the PV confidence capacity will be more affected. This can be understood from Fig. 6 as the time interval increases, the similarity correlation index of PV-load becomes higher and higher, resulting in a larger PV confidence capacity value; the optimal time step for PV confidence capacity evaluation varies with PV penetration In Figure 9, the effect of time step on the credit rating of photovoltaic power generation capacity decreases with the increase of photovoltaic penetration. In practical applications, a larger time scale can be selected for PV confidence capacity assessment to improve PV penetration, speed up computation, and maintain computational accuracy. The confidence capacity for a given time step can be evaluated using the trained artificial neural network shown in Figure 3. This process is repeated for different time steps to obtain the confidence capacity sequence, which can be used to determine the optimal time interval. This process is repeated for different time steps to obtain the confidence capacity sequence, which can be used to determine the optimal time interval. Table 2 gives the reference values of Δt _opt at different PV permeability at η=0.1687. The detailed results of PV confidence capacity assessment using the proposed empirical model are presented in Fig. 10.

Table 2 Optimal time steps for PV confidence capacity evaluation under different PV penetration rates

After the simulation and verification, it can be found that the large-scale grid-connected photovoltaic confidence capacity evaluation method based on the empirical model proposed in the present invention can avoid the use of complex and time-consuming SMC calculations, and can be used at any given photovoltaic permeability, time step and Confidence capacity accurately estimated under similar conditions related to PV load.

Embodiment 2: Based on the same inventive concept, the present invention also provides a system for determining the confidence capacity of a grid-connected photovoltaic system, including:

The acquisition module is used to obtain the PV penetration rate, time step and PV load related similarity index;

a determination module, used for bringing the photovoltaic penetration rate, time step and photovoltaic load related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

Wherein, the empirical model is obtained by using artificial neural network to train the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index.

In an embodiment, the system further includes a training module; the training module is specifically used for:

Set multiple photovoltaic capacities;

For each photovoltaic capacity, simulate under different time steps to obtain the corresponding photovoltaic output power distribution;

A sample set is formed based on all the time steps corresponding to each photovoltaic capacity and the photovoltaic output power distribution;

Build artificial neural networks based on different PV penetration rates, time steps and PV load-related similarity values;

The artificial neural network is trained based on the sample set, and the mapping relationship between confidence capacity and photovoltaic permeability, time step and photovoltaic load related similarity index is obtained.

In the embodiment, the mapping relationship between the confidence capacity in the empirical model and the photovoltaic permeability, the time step and the photovoltaic load related similarity index is shown in the following formula:

κ _CC =g(r,Δt,η)

In the formula: κ _CC is the confidence capacity of the photovoltaic system; r is the photovoltaic penetration rate; Δt is the time step; η is the photovoltaic load correlation similarity index in time order.

In the embodiment, the photovoltaic load-related similarity index η is shown in the following formula:

η=v× _γC

In the formula: v is the ramp rate index of photovoltaic output power according to time change, γ _C is the correlation between photovoltaic output power and load distribution.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flows of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

The above are only examples of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention are included in the application for pending approval of the present invention. within the scope of the claims.

Claims (10)

1. A method for determining confidence capacity of a grid-connected photovoltaic system is characterized by comprising the following steps:

obtaining photovoltaic permeability, time step and photovoltaic load related similarity indexes;

substituting the photovoltaic permeability, the time step and the photovoltaic load related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

the empirical model is obtained by training the mapping relation between the confidence capacity and the photovoltaic permeability, the time step and the photovoltaic load related similarity index by using an artificial neural network.

2. The method of claim 1, wherein the training of the empirical model comprises:

setting a plurality of photovoltaic capacities;

for each photovoltaic capacity, simulating under different time step lengths to obtain corresponding photovoltaic output power distribution;

forming a sample set based on all time step lengths corresponding to all photovoltaic capacities and photovoltaic output power distribution;

establishing an artificial neural network based on different photovoltaic permeabilities, time step lengths and photovoltaic load related similarity values;

and training the artificial neural network based on the sample set to obtain a mapping relation between the confidence capacity and the photovoltaic permeability, the time step length and the photovoltaic load related similarity index.

3. The method of claim 2, wherein the empirical model maps confidence capacity with photovoltaic permeability, time step size, and photovoltaic load related similarity index as follows:

κ_CC＝g(r,Δt,η)

in the formula: kappa_CCIs the confidence capacity of the photovoltaic system; r is photovoltaic permeability; Δ t is the time step; η is the chronological photovoltaic-load related similarity index.

4. The method of claim 3, wherein the photovoltaic-load related similarity index η is represented by the following formula:

η＝v×γ_C

in the formula: v is a ramp rate indicator of the power of the photovoltaic output as a function of time, gamma_CIs the dependence of the photovoltaic output power on the load distribution.

5. The method of claim 4, wherein the power of the photovoltaic output is ramped in a time-varying rate indicator v, as shown in the following equation:

in the formula: n represents the number of photovoltaic output units;

is a load time sequence;

is the absolute maximum fluctuation of the photovoltaic output and load in the time step Δ t;

a normalized ramp rate representing a time series of photovoltaic output power;

representing the corresponding photovoltaic capacity of the (i + 1) th photovoltaic output unit at the time step delta t;

representing the corresponding photovoltaic capacity of the ith photovoltaic output unit at the time step delta t;

representing the normalized load ramp rate over time T; l is_(i+1)ΔtRepresenting the fluctuation of the photovoltaic output and load of the (i + 1) th photovoltaic output unit at the time step delta t; l is_iΔtRepresenting the fluctuation of the photovoltaic output and load of the ith photovoltaic output unit at the time step deltat.

6. The method of claim 2, wherein the artificial neural network is a back propagation neural network.

7. A grid-connected photovoltaic system confidence capacity determination system is characterized by comprising:

the acquisition module is used for acquiring photovoltaic permeability, a time step and a photovoltaic load related similarity index;

the determining module is used for substituting the photovoltaic permeability, the time step and the photovoltaic load related similarity index into a pre-trained empirical model to determine the confidence capacity of the photovoltaic system;

the empirical model is obtained by training the mapping relation between the confidence capacity and the photovoltaic permeability, the time step and the photovoltaic load related similarity index by using an artificial neural network.

8. The system of claim 7, further comprising a training module; the training module is specifically configured to:

setting a plurality of photovoltaic capacities;

for each photovoltaic capacity, simulating under different time step lengths to obtain corresponding photovoltaic output power distribution;

forming a sample set based on all time step lengths corresponding to all photovoltaic capacities and photovoltaic output power distribution;

establishing an artificial neural network based on different photovoltaic permeabilities, time step lengths and photovoltaic load related similarity values;

and training the artificial neural network based on the sample set to obtain a mapping relation between the confidence capacity and the photovoltaic permeability, the time step length and the photovoltaic load related similarity index.

9. The system of claim 8, wherein the empirical model maps confidence capacity with photovoltaic permeability, time step size, and photovoltaic load related similarity index as follows:

κ_CC＝g(r,Δt,η)

in the formula: kappa_CCIs the confidence capacity of the photovoltaic system; r is photovoltaic permeability; Δ t is the time step; η is a chronological photovoltaic load related similarity index.

10. The method of claim 9, wherein the photovoltaic load-related similarity index η is represented by the following formula:

η＝v×γ_C

in the formula: v is a ramp rate indicator of the power of the photovoltaic output as a function of time, gamma_CIs the dependence of the photovoltaic output power on the load distribution.

Images