Regional photovoltaic anti-reflux control method considering adjustable user risk

Abstract

The invention relates to the technical field of photovoltaic countercurrent control, and particularly discloses a regional photovoltaic countercurrent prevention control method considering adjustable user risks, and the method comprises the steps: carrying out the real-time prediction of the power load of a transformer region based on a Tide algorithm; determining power control values of distributed photovoltaic output at different moments by combining historical load data and predicted load data of the transformer area; and establishing an evaluation index of the countercurrent risk of the power distribution network area, and determining the increase or decrease of the photovoltaic reinforcement according to a risk threshold in the evaluation index. According to the invention, a Tide model based on MLP is adopted to carry out transformer area load prediction; according to the invention, the photovoltaic output power limit value can be dynamically adjusted, and the countercurrent risk caused by too high photovoltaic output is avoided; four different photovoltaic output control strategies are designed, and the optimal scheme can be flexibly selected according to the real-time change of the transformer area load. A reliable risk assessment basis is provided for safe operation of a power distribution network area by defining a countercurrent risk assessment index and calculating a countercurrent risk value by using a kernel density function.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic backflow control, and in particular to a regional photovoltaic backflow prevention control method taking into account adjustable user risks. Background Art

[0002] Photovoltaic power generation has become a focus because it is clean, low-carbon and renewable, especially distributed photovoltaics, which have attracted great attention due to their rich application scenarios and development potential. Although distributed photovoltaics are easy to build, high-proportion access brings challenges to the stable operation of the power grid.

[0003] First, the output of photovoltaic power generation is greatly affected by weather and climate, and has strong randomness and volatility, and the output is often staggered with the user's power load. Secondly, distributed photovoltaic power generation usually has the characteristics of small installed capacity, large number of individuals, and wide distribution, so it is difficult to adopt the traditional centralized dispatching method to coordinate and optimize the control of the distribution network. In addition, the grid connection of distributed photovoltaic power generation often leads to increased node voltage, and the mismatch between photovoltaic output and power load further increases the risk of voltage over-limit and reverse overload. These problems have brought challenges to the healthy and stable operation of the distribution network. It is particularly noteworthy that there are significant differences in the requirements for reverse flow risk control and the absorption capacity near the reverse flow in different regions, which makes it impossible for a unified control strategy to meet the needs. The grid structure, user power characteristics, and photovoltaic resource distribution of each region are different, resulting in the need for stricter reverse flow control in some areas, while in other areas there may be higher absorption capacity. Therefore, when designing the control strategy, it is necessary to take into account the differences between regions to ensure the safe and stable operation of the distribution network.

[0004] Secondly, although a large number of technical studies have focused on the problem of backflow in distributed power grids and adopted a series of methods to design backflow prevention, so far, no relevant technical solutions have taken into account the needs of the user side and established the interactive function between user risks and actual solutions. When different risk preferences appear on the demand side, it is necessary to adjust the intensity of backflow prevention in real time through algorithm design to ensure that the customer's demand for electricity economy is met, so as to maximize energy utilization and optimize overall economic benefits.

[0005] Therefore, a regional photovoltaic backflow prevention control method that takes into account the risks of adjustable users has become an urgent problem to be solved. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a regional photovoltaic anti-backflow control method that takes into account the risks of adjustable users. The method mainly ensures that the photovoltaic output matches the local load, and uses the data collected by the backflow collection device and the photovoltaic local power generation data to perform real-time regulation of the photovoltaic power limit. No backflow occurs, and the photovoltaic output is increased on the basis of no backflow.

[0007] In order to solve the above technical problems, the present invention provides a technical solution: a regional photovoltaic backflow prevention control method considering adjustable user risks, comprising the following steps:

[0008] S1. Real-time prediction of power load in the substation based on Tide algorithm;

[0009] S2. Determine the power control value of distributed photovoltaic output at different times by combining the historical load data and predicted load data of the substation area;

[0010] S3. Establish assessment indicators for reverse flow risk in distribution network substations, and determine the increase or decrease of photovoltaic power generation based on the risk threshold in the assessment indicators.

[0011] Furthermore, in step S1, the prediction method is specifically as follows:

[0012] The past and future covariates are projected and combined with the historical data and static attributes of the time series, and then the combined data is encoded through a dense encoder containing a multi-layer perceptron;

[0013] After the original information is encoded, the encoded information is converted into a predicted time series through a dense MLP decoder;

[0014] The prediction result is obtained through the temporal decoder.

[0015] Furthermore, the r-dimensional dynamic covariate of the time series y at time t is expressed as x t ∈R r ; Use a to represent the static properties of the time series;

[0016] The predictor is described by a function f that is derived from the historical data y of the time series 1:L , dynamic covariate x 1:L+H , static attribute a is mapped to accurate prediction of the future:

[0017]

[0018] The method of encoding the combined data through a dense encoder containing a multi-layer perceptron is as follows:

[0019]

[0020] The method of converting the encoded information into a predicted time series through a dense MLP decoder is as follows:

[0021] The decoder first uses the code e to generate an H×p output vector g, where p is the decoder output dimension;

[0022] The vector g is reshaped into a matrix D of dimension d×H;

[0023] The tth column d of the matrix D t Expressed as the decoded vector at time t:

[0024] g = Decoder(e);

[0025] D = Reshape(g);

[0026] The prediction result is obtained through the time series decoder, which introduces a direct line from the covariate at time L+t to the predicted value at time L+t, as follows:

[0027]

[0028] Furthermore, the method for determining the distributed photovoltaic output power limit at different times is as follows:

[0029] The predicted load at the current moment is used as the power limit of distributed photovoltaic output; that is, the power load at the current moment is predicted in real time based on the Tide prediction model, and then this predicted value is used as the power limit of distributed photovoltaic output.

[0030] Furthermore, the method for determining the distributed photovoltaic output power limit at different times includes:

[0031] The minimum load in the previous i minutes is taken as the power limit of distributed photovoltaic output, where i∈{1, 2, 3, 4, 5}.

[0032] Furthermore, the method for determining the distributed photovoltaic output power limit at different times is as follows:

[0033] The minimum value of the lowest load in the previous i minutes and the predicted load at the current moment is taken as the power limit of the distributed photovoltaic output, where i∈{1, 2, 3, 4, 5}.

[0034] Furthermore, the method for determining the distributed photovoltaic output power limit at different times is as follows:

[0035] The minimum value of the lowest load in the previous i minutes and the predicted load in the next j minutes is taken as the power limit of the distributed photovoltaic output, where i, j∈{1, 2, 3}.

[0036] Furthermore, in step S3, the backflow risk R is defined, and its calculation formula is as follows:

[0037]

[0038] Among them, threshold is the reserved risk threshold, that is, the predicted control power minus the actual control power, and max is the maximum reverse flow power value;

[0039] f(x) is the kernel density function. When calculating the backflow risk R, it is necessary to first determine the kernel density function f(x) of the backflow power data;

[0040] The formula for kernel density estimation is:

[0041]

[0042] Among them, n is the number of data points; h is the bandwidth parameter, which determines the degree of smoothing; K is the kernel function, which is expressed as:

[0043]

[0044] According to the K(u) formula, the predicted power function is divided into several intervals according to the hour, so as to determine the prediction error kernel density function of the predicted control power in each interval; then according to the reverse flow risk R, the reserved risk threshold threshold for each period is obtained; the numerical method is used to solve it, and the solution formula is:

[0045]

[0046] threshold=max-xh;

[0047] h = 0.0001;

[0048] For the calculated reserved risk threshold, if the value exceeds the set risk threshold, it is necessary to reduce the photovoltaic output to reduce the reverse flow risk; if the value is lower than the set risk threshold, the photovoltaic output is increased to maximize the photovoltaic utilization rate.

[0049] Furthermore, simulation sampling is performed on the user group, and the actual backflow probability in the observation data set should be less than or equal to the backflow risk tolerable probability preset by the user, that is, the backflow controllability rate; thus, the performance indicator is set: the backflow controllability rate>99.99%.

[0050] The advantages of the present invention compared with the prior art are:

[0051] 1. The present invention adopts a time series dense encoder-decoder (Tide) model based on a multi-layer perceptron (MLP) to predict load in the substation area. The Tide model can not only process complex time series data, but also effectively use dynamic covariates and static attributes for long-term prediction, thereby improving the accuracy and stability of load prediction.

[0052] 2. The present invention can dynamically adjust the photovoltaic output power limit to avoid the risk of reverse flow caused by excessive photovoltaic output. At the same time, through real-time regulation, it ensures the maximum utilization of photovoltaic power generation, reduces the phenomenon of abandoned light, and improves the economic benefits of photovoltaic power generation.

[0053] 3. The present invention designs four different photovoltaic output control strategies, which can flexibly select the optimal solution according to the real-time changes in the load of the substation. This diversified control strategy can not only cope with various load fluctuations, but also significantly improve the penetration rate and utilization efficiency of photovoltaic power generation.

[0054] 4. The present invention defines the reverse flow risk assessment index and uses the kernel density function to calculate the reverse flow risk value, thus providing a reliable risk assessment basis for the safe operation of the distribution network. The assessment system can provide early warning of possible reverse flow risks and ensure the safety and stability of the distribution network.

[0055] 5. The present invention enables users to adjust photovoltaic output and load according to actual needs through accurate load prediction and flexible control strategies, ensuring electricity safety and economic benefits under various load fluctuations, thereby achieving effective control of user risks. The time series dense encoder-decoder (Tide) model based on multi-layer perceptron (MLP) can provide high-precision load prediction, allowing users to understand future electricity demand in advance and prepare in advance to avoid the risks caused by sudden increase or decrease in electricity load. The four different photovoltaic output control strategies designed allow users to choose the most suitable strategy for adjustment according to actual conditions, such as reducing photovoltaic output during peak electricity consumption to avoid overload risks, and increasing photovoltaic output during low electricity consumption to improve utilization. Through the real-time data monitoring system, users can grasp the real-time situation of photovoltaic power generation and load at any time, discover and respond to abnormal conditions in time, and ensure the stability and safety of the power system. In addition, through the backflow risk assessment system, users can warn of possible backflow risks in advance and take preventive measures to avoid power grid safety problems caused by backflow. Ultimately, through accurate load forecasting and flexible control strategies, users can maximize the use of photovoltaic power generation, reduce the phenomenon of abandoned light, improve the economic benefits of photovoltaic power generation, while avoiding economic losses caused by excessive or low photovoltaic output, and optimize electricity costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a system framework diagram of a regional photovoltaic backflow prevention control method considering adjustable user risks according to the present invention.

[0057] Figure 2 This is a typical Tide algorithm process. DETAILED DESCRIPTION

[0058] Various exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0059] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0060] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0061] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0062] The following is a further detailed description of a regional photovoltaic backflow prevention control method considering adjustable user risks of the present invention in conjunction with the accompanying drawings.

[0063] Combined with Figure 1-2 The specific implementation process of the regional photovoltaic backflow prevention control method considering adjustable user risks of the present invention is as follows:

[0064] Firstly, the real-time prediction of the power load in the substation is realized based on the Tide algorithm, and the power control value of the distributed photovoltaic output at different times is determined by combining the historical load data and the predicted load data of the substation. Secondly, the evaluation index of the reverse flow risk of the distribution network substation is established to provide a reference for the risk assessment and control of the distribution network substation containing distributed photovoltaic.

[0065] The present invention is implemented by the following specific method:

[0066] General Overview:

[0067] This method mainly ensures that the photovoltaic output matches the local load, and uses the data collected by the reverse flow collection device and the photovoltaic local power generation data to conduct real-time regulation of the photovoltaic power limit. The first goal is to prevent reverse flow, and the second goal is to increase the photovoltaic output on the basis of no reverse flow.

[0068] The input data are the active power and reverse meter power of each photovoltaic inverter over a period of history, the real-time power, rated power, and reverse meter power of each photovoltaic inverter, and the output data are the control strategies of each photovoltaic inverter.

[0069] Detailed introduction:

[0070] 1. Holiday load forecasting based on Tide algorithm

[0071] In order to accurately determine the power control value of photovoltaic output at different times, it is necessary to first accurately estimate the power load of the distribution network area. The time-series dense encoder (Tide) is an encoder-decoder model that relies on a multi-layer perceptron (MLP) and is considered to be efficient in long-term time series prediction. Figure 2 The typical Tide algorithm flow is shown.

[0072] For long-term time series forecasting, the core issue is to use the historical data y of the time series 1:L Conduct future data L+1:L+H In many forecasting scenarios, there may be covariates of the time series, which are usually known in advance, including global covariates and time series-specific covariates.

[0073] In the present invention, the r-dimensional dynamic covariate of the time series y at time t is expressed as x t ∈R r In addition, a is used to represent the static properties of the time series, that is, the characteristics that do not change over time. In this way, the predictor can be described by a function f, which is derived from the historical data y of the time series. 1:L , dynamic covariate x 1:L+H , static attribute a is mapped to accurate prediction of the future:

[0074]

[0075] The implementation of the flowchart of function f is described in detail below. First, the past and future covariates are projected and combined with the historical data and static properties of the time series, and then the combined data is encoded by a dense encoder containing a multi-layer perceptron:

[0076]

[0077] After encoding the original information, the algorithm converts the encoded information into a predicted time series through a dense MLP decoder. The decoder first uses the encoding e to generate an H×p output vector g, where p is the decoder output dimension. Then, the vector g is reshaped into a matrix D of dimension d×H. The tth column d of the matrix D is t It can be expressed as the decoded vector at time t:

[0078] g = Decoder(e);

[0079] D = Reshape(g);

[0080] The final prediction is obtained through a time series decoder, which introduces a direct path from the covariate at time L+t to the predicted value at time L+t, which is particularly useful when the covariate has a significant impact on the actual value at some specific time:

[0081]

[0082] The input data of the present invention is the historical data of the previous 60 minutes, and the output data is the load power of the next minute of the sequence group. In this way, the power load prediction model of the distribution network based on the Tide algorithm can accurately predict the power load of the distribution network at different time points to meet the needs of future photovoltaic output power control.

[0083] 2. Best control strategy for backflow prevention

[0084] The reverse heavy overload problem is caused by the output level of distributed photovoltaics being higher than the power load of the substation. Therefore, how to determine the output level of distributed photovoltaics based on the power load of the substation is the key to solving the reverse heavy overload problem. However, due to the time series changes in the power load of power users, directly determining a unified photovoltaic power control value may lead to excessive abandonment of light during peak power periods, resulting in waste. In summary, the objectives of the present invention are mainly the following two points: one is to ensure that there is no reverse flow in the distribution network substation by achieving real-time regulation of photovoltaic power limits; the second is to increase the distributed photovoltaic output as much as possible without reverse flow, thereby reducing abandonment and increasing the penetration rate of renewable energy.

[0085] In order to determine the power limit of distributed photovoltaic output at different times, the present invention proposes four different control strategies to find the best control scheme for distributed photovoltaic output. The specific contents of the four schemes are as follows:

[0086] Option 1: Forecasting Load

[0087] The predicted load at the current moment is used as the power limit of distributed photovoltaic output. This solution predicts the current power load in real time based on the Tide prediction model, and then uses this predicted value as the power limit of distributed photovoltaic output. The advantage of this method is that it is simple and direct, and can quickly respond to changes in the current load. However, since the prediction model may have certain errors, when the load changes rapidly, the adjustment of the photovoltaic output limit value may not be accurate enough, which may cause certain risks.

[0088] Solution 2: min{minimum load in the previous {i} minutes}, where i∈(1, 2, 3, 4, 5)

[0089] The lowest load in the previous i minutes is used as the power limit of distributed photovoltaic output, where i∈{1, 2, 3, 4, 5)}. In this scheme, the lowest load value in the previous i minutes is selected as the power limit of the current photovoltaic output. The reason for this is that by adopting the recent lowest load value, the risk of reverse flow caused by extremely low instantaneous load can be avoided to a certain extent. This method relies more on historical load data and can better cope with short-term fluctuations in load, but when the load is relatively stable or the trend is rising, it may be conservative, resulting in a low limit setting.

[0090] Scheme 3: min{minimum load in the previous {i} minutes, predicted load}, where i∈(1, 2, 3, 4, 5)

[0091] The minimum value of the lowest load in the previous i minutes and the predicted load at the current moment is used as the power limit of distributed photovoltaic output, where i∈{1, 2, 3, 4, 5)}. This scheme comprehensively considers the historical load and the current predicted load, and selects the smaller value as the power limit of photovoltaic output. The advantage of this approach is that it can avoid the error influence of a single predicted load model and improve the robustness of decision-making with the help of historical load data information. When the load fluctuates greatly, this scheme can more effectively prevent the risk of reverse flow, and also reduce the situation of abandoned light under stable load.

[0092] Scheme 4: min{minimum load in the first {i} minutes, predicted load in the next {j} minutes}; where i, j∈(1, 2, 3)

[0093] The minimum value of the lowest load in the previous i minutes and the predicted load in the future j minutes is used as the power limit of the distributed photovoltaic output, where i, j∈{1, 2, 3}. This scheme not only combines the historical load and the current predicted load, but also introduces the load forecast information in the short term in the future. By using the minimum value of the lowest load in the i minute and the minimum value of the predicted load in the j minute as the limit, it is possible to respond more accurately to the upcoming load changes in the future. The advantage of this method is that when the load may rise rapidly, it can be regulated in advance to avoid overly conservative limit settings and improve the effective utilization of photovoltaic output. However, this method has higher requirements for the load forecasting model and needs to ensure the accuracy of the forecast.

[0094] For different distribution network areas, the power limits of distributed photovoltaic output under four schemes can be calculated through historical photovoltaic data and power load data, and then compared with the actual power load conditions in the area to obtain the optimal distributed photovoltaic output control plan.

[0095] 3. Distribution network area reverse flow risk assessment indicators

[0096] In order to evaluate the reverse flow risk of distribution network substations under different control strategies, the present invention defines the reverse flow risk R, which is calculated as follows:

[0097]

[0098] Among them, threshold is the reserved risk threshold, that is, the predicted control power minus the actual control power, and max is the maximum reverse flow power value.

[0099] f(x) is the kernel density function (KDE, Kernel Density Estimation), which is an important tool for estimating the probability density distribution of data. When calculating the risk factor R, it is necessary to first determine the probability density function (f(x)) of the reverse power data.

[0100] The formula for kernel density estimation is:

[0101]

[0102] Among them, n is the number of data points; h is the bandwidth parameter, which determines the degree of smoothing; K is the kernel function, and the commonly used kernel function is the Gaussian kernel, which is expressed as:

[0103]

[0104] According to the above formula, the predicted power function is divided into several intervals according to the hour, so as to determine the prediction error kernel density function of the predicted control power in each interval; then according to the user's risk tolerance R, the reserved risk threshold threshold for each period is obtained. Since the original function of the kernel density cannot be obtained directly, a numerical method can be used to solve it. The solution formula is:

[0105]

[0106] threshold=max-xh;

[0107] h = 0.0001;

[0108] For the calculated reserved risk threshold, if the value exceeds the set risk threshold, it is necessary to reduce the photovoltaic output to reduce the reverse flow risk; if the value is lower than the set risk threshold, the photovoltaic output can be increased to maximize the photovoltaic utilization rate.

[0109] Finally, to understand the comprehensiveness of the algorithm evaluation, the following performance indicators were added: backflow controllability rate > 99.99%.

[0110] Through the above method, the user group is simulated and sampled, and the actual reverse flow probability in the observed data set should be less than or equal to the reverse flow risk tolerance probability preset by the user, that is, the reverse flow controllability rate. The strategy of photovoltaic output under risk demand is further adjusted by setting the threshold. The set 99.99% threshold can ensure that the photovoltaic system can meet the risk tolerance requirements of most users, thereby improving the stability and safety of the system.

[0111] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A regional photovoltaic backflow prevention control method considering adjustable user risks, characterized in that: The following steps are included S1. Real-time prediction of power load in the substation based on Tide algorithm; S2. Determine the power control value of distributed photovoltaic output at different times by combining the historical load data and predicted load data of the substation area; S3. Establish assessment indicators for reverse flow risk in distribution network substations, and determine the increase or decrease of photovoltaic power generation based on the risk threshold in the assessment indicators.

2. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 1, characterized in that: In step S1, the prediction method is as follows: The past and future covariates are projected and combined with the historical data and static attributes of the time series, and then the combined data is encoded through a dense encoder containing a multi-layer perceptron; After the original information is encoded, the encoded information is converted into a predicted time series through a dense MLP decoder; The prediction result is obtained through the temporal decoder.

3. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 2, characterized in that: For the time series y at time t, the r-dimensional dynamic covariate is represented by x t ∈R r ; Use a to represent the static properties of the time series; The predictor is described by a function f that is derived from the historical data y of the time series 1:L , dynamic covariate x 1:L+H , static attribute a is mapped to accurate prediction of the future: The method of encoding the combined data through a dense encoder containing a multi-layer perceptron is as follows: The method of converting the encoded information into a predicted time series through a dense MLP decoder is as follows: The decoder first uses the code e to generate an H×p output vector g, where p is the decoder output dimension; The vector g is reshaped into a matrix D of dimension d×H; The tth column d of the matrix D t Expressed as the decoded vector at time t: g = Decoder(e); D = Reshape(g); The prediction result is obtained through the time series decoder, which introduces a direct line from the covariate at time L+t to the predicted value at time L+t, as follows:

4. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 3, characterized in that: The method for determining the distributed photovoltaic output power limit at different times is as follows: The predicted load at the current moment is used as the power limit of distributed photovoltaic output; that is, the power load at the current moment is predicted in real time based on the Tide prediction model, and then this predicted value is used as the power limit of distributed photovoltaic output.

5. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 3, characterized in that: Methods for determining the distributed photovoltaic output power limit at different times include: The minimum load in the previous i minutes is taken as the power limit of distributed photovoltaic output, where i∈{1, 2, 3, 4, 5}.

6. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 3, characterized in that: The method for determining the distributed photovoltaic output power limit at different times is as follows: The minimum value of the lowest load in the previous i minutes and the predicted load at the current moment is taken as the power limit of the distributed photovoltaic output, where i∈{1, 2, 3, 4, 5}.

7. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 3, characterized in that: The method for determining the distributed photovoltaic output power limit at different times is as follows: The minimum value of the lowest load in the previous i minutes and the predicted load in the next j minutes is taken as the power limit of the distributed photovoltaic output, where i, j∈{1, 2, 3}.

8. A regional photovoltaic backflow prevention control method considering adjustable user risks according to any one of claims 4 to 7, characterized in that: In step S3, the backflow risk R is defined, and its calculation formula is as follows: Among them, threshold is the reserved risk threshold, that is, the predicted control power minus the actual control power, and max is the maximum reverse flow power value; f(x) is the kernel density function. When calculating the backflow risk R, it is necessary to first determine the kernel density function f(x) of the backflow power data; The formula for kernel density estimation is: Among them, n is the number of data points; h is the bandwidth parameter, which determines the degree of smoothing; K is the kernel function, which is expressed as: According to the K(u) formula, the predicted power function is divided into several intervals according to the hour, so as to determine the prediction error kernel density function of the predicted control power in each interval; then according to the reverse flow risk R, the reserved risk threshold threshold for each period is obtained; the numerical method is used to solve it, and the solution formula is: threshold=max-xh; h＝0.0001； For the calculated reserved risk threshold, if the value exceeds the set risk threshold, it is necessary to reduce the photovoltaic output to reduce the reverse flow risk; if the value is lower than the set risk threshold, the photovoltaic output is increased to maximize the photovoltaic utilization rate.

9. A regional photovoltaic backflow prevention control method considering adjustable user risks according to claim 8, characterized in that: Perform simulation sampling on the user group and observe that the actual backflow probability in the data set should be less than or equal to the backflow risk tolerance probability preset by the user, that is, the backflow controllability rate; thus, the performance indicator is set: the backflow controllability rate>99.99%.

Images