Inverter intelligent regulation and control method integrating power dynamic calculation and abnormal value elimination

Abstract

The invention relates to the technical field of power regulation and control, and discloses an inverter intelligent regulation and control method integrating power dynamic calculation and abnormal value elimination, and the method comprises the steps: calculating an adjustable coefficient of each inverter based on real-time power; calculating the total power reduction quantity and the total adjustable power of all inverters in the photovoltaic transformer area, and distributing initial reduction power for each inverter; calculating undistributed power based on the initial reduction power, and distributing reduction power to the inverters again based on the undistributed power to obtain planned reduction power of each inverter; inputting the planned reduction power into the optimization model to obtain final reduction power, and performing power reduction on each inverter according to the final reduction power; and judging whether the inverter subjected to power reduction operates abnormally or not by using a local outlier factor algorithm, and if the inverter operates abnormally, intelligently recovering the power of the inverter to be rated power. The power generation power of the inverter is intelligently regulated and controlled, the inverter is prevented from being in an overload state for a long time, and the operation stability of the photovoltaic transformer area is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power regulation, in particular to an intelligent inverter regulation method combining power dynamic calculation and abnormal value elimination. BACKGROUND

[0002] With the large-scale popularization and high proportion of photovoltaic power generation technology in distribution areas, the operation paradigm of distribution networks is undergoing a fundamental change. Traditionally, the terminal area is a passive receiving and distribution end of electric energy, but now it has become a hybrid node with self-generation and power utilization capability. This structural change has caused a significant deviation in its operating characteristics. During the day when the light resources are abundant, especially during the noon when the load is relatively low, the total amount of photovoltaic system power generation is likely to exceed the local load demand, resulting in serious power back feeding and voltage rise, causing the area voltage to cross the line, the power quality to deteriorate (such as voltage fluctuation and harmonic pollution), and the primary equipment (such as transformers and lines) to face the risk of long-term overload. The traditional power utilization information acquisition system and line loss calculation model also deviate significantly due to the presence of two-way flow, making the fine operation and management of the area face severe challenges.

[0003] To cope with the core contradiction of overvoltage and overload induced by high proportion of photovoltaic access, and to improve the consumption capacity and operation resilience of the distribution system, the industry generally adopts the strategy of issuing regulation instructions to photovoltaic inverters to limit their power generation. However, the current control method has some limitations. First, its regulation logic is mostly based on a simple judgment of whether the total power generation of the area exceeds a certain preset threshold, and then a general power regulation parameter is issued to the inverter. This control method causes the regulation instruction to deviate from the actual system demand, and the actual reduced power often deviates greatly from the expected target. Second, the existing regulation strategy lacks adaptability. Since photovoltaic power generation and inverter power have significant randomness and volatility, fixed regulation parameters cannot match them, forcing the operation personnel to make frequent strategy adjustments and parameter updates every day, resulting in low management efficiency. SUMMARY

[0004] The purpose of the present application is to intelligently regulate the power generation of inverters, avoid the total power generation of photovoltaic areas exceeding the local load demand, and cause the inverter to be in an overload state for a long time, improve the operation stability of photovoltaic areas, and provide an intelligent inverter regulation method combining power dynamic calculation and abnormal value elimination.

[0005] To achieve the above-mentioned application purpose, the embodiments of the present application provide the following technical solutions:

[0006] The intelligent inverter regulation method combining power dynamic calculation and abnormal value elimination comprises the following steps:

[0007] Step 1, obtain the real-time power of the inverter, and calculate the adjustable coefficient of each inverter based on the real-time power;

[0008] Step 2, calculate the total power reduction and total adjustable power of all inverters in the photovoltaic station area, and allocate initial reduction power to each inverter;

[0009] Step 3, calculate the unallocated power based on the initial reduction power, and allocate reduction power to the inverter again based on the unallocated power to obtain the planned reduction power of each inverter;

[0010] Step 4, input the planned reduction power into the optimization model to obtain the final reduction power, and reduce the power of each inverter according to the final reduction power;

[0011] Step 5, fine-tune the inverter power;

[0012] Step 6, use the local outlier factor algorithm to determine whether the inverter after power reduction is running abnormally, and if so, intelligently restore the power of the inverter to the rated power.

[0013] The calculation formula of the adjustable coefficient in step 1 is:

[0014] wherein, is the weight of the i-th inverter, , P rated,i is the rated power of the i-th inverter; n is the number of inverters in the photovoltaic station area; and α and β are adjustable coefficient weight factors.

[0015] In step 2, the total power reduction of all inverters in the photovoltaic station area is calculated as:

[0016] wherein, is the total power reduction; P over is the output overrun value; P total is the rated output power of all inverters; and P max is the upper limit value of safe power operation of the photovoltaic station area.

[0017] In step 2, the initial reduction power allocated to each inverter is:

[0018] wherein, is the initial reduction power of the i-th inverter;

[0019] The initial reduction power allocated to all inverters is Store to the power reduction queue U, calculate every minute And update the power reduction queue U.

[0020] In step 3, the unallocated power based on the initial power reduction calculation is:

[0021] Wherein, P undistributed Is the unallocated power of the photovoltaic station area.

[0022] In step 3, the power reduction is allocated to the inverter again based on the unallocated power, and the planned power reduction of each inverter is:

[0023] Delete the inverter with real-time power P new,i 0 in the power reduction queue U, get the effective inverter list B, and allocate the power reduction to the inverter in the effective inverter list B again:

[0024] Wherein, Is the power reduction allocated to the i-th inverter in the effective inverter list B for the t-th time; t≥2; P undistributed_t Is the unallocated power of the photovoltaic station area after t times of power reduction allocation to the inverter in the effective inverter list B; x is the number of inverter deleted from the power reduction queue U, x≥0; i=1,2,...,n-x;

[0025] Construct the power loss model L i of the i-th inverter:

[0026]

[0027] Wherein, a is the conversion efficiency of the i-th inverter; b is the current price; c is the starting loss of the i-th inverter;

[0028] Calculate the current loss L i_now of the i-th inverter in the effective inverter list B, solve the value of i_now when the current loss L is minimum:

[0029] Wherein, Indicates the t-order derivative of the power loss model L i ; the t value corresponding to the minimum value of L i_now is obtained by derivation;

[0030] The power reduction calculated t times and the initial power reduction Add up, update the planned curtailment power .

[0031] In step 4, the planned curtailment power Input the optimization model, and the final curtailment power obtained is:

[0032] Define the objective function and two constraint conditions of the optimization model:

[0033] Objective function ;

[0034] Power balance constraint condition ;

[0035] Power non-negative constraint condition ;

[0036] Construct the Lagrange function:

[0037] Wherein, L is the Lagrange function; C is the marginal regulation cost;

[0038] Take the partial derivative of the Lagrange function and set it to 0:

[0039] Wherein, L is the Lagrange function;

[0040] Take the partial derivative of and set it to 0:

[0041] There is:

[0042] That is:

[0043] Finally get:

[0044] Wherein, is the final curtailment power of the i-th inverter;

[0045] The final curtailment power allocated by all inverters is stored in the curtailment power queue U, and the final curtailment power is calculated every minute And update the curtailment power queue U.

[0046] The step 6 specifically comprises the following steps:

[0047] Read the inverter's real-time power P every 5 seconds. new,i and P new,i Store in queue C;

[0048] Based on queue C, maintain recent historical data of fixed length L, and set the sliding window to W=[P t-L ,P t-L+1 ,...,P t-1 For the data point P newly added to queue C new,i Calculate the two dynamic quantiles of window W, namely the upper quantile Q. 1-α and lower quantile Q α If P new,i α or P new,i Q 1-α Then the data point P new,i It was identified as an outlier and the data point was directly removed from queue C;

[0049] Otherwise, the local outlier factor algorithm is used to identify whether the data point is abnormal. If it is abnormal, the data point is removed from queue C.

[0050] After removing outliers from queue C, calculate the average real-time power in queue C. :

[0051] Where N is the storage capacity of queue C;

[0052] like Adjust the real-time power of the inverter back to its rated power, where P achieve,i For the desired power output of the i-th inverter after reduction, we have:

[0053] P current,i The power of the i-th inverter before it was reduced.

[0054] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention calculates the power reduction of each inverter in three stages. The first stage is to calculate the initial power reduction, which can be done simply based on the total power reduction of the photovoltaic area. However, considering that the greater the power reduction of the inverter, the greater the loss and economic loss will be, the power of the inverter cannot be blindly reduced. Therefore, the second stage, while increasing the power reduction of the inverter, uses the power loss model L... i ​The constraint guarantees that the loss of the inverter, the economic loss is reduced, the planned power reduction is calculated; and considering the operation stability of the inverter after the power reduction, the third stage optimizes the power reduction distribution problem by constructing an optimization model, and calculates the final power reduction to improve the stability of the inverter. After the power reduction of the inverter according to the final power reduction, the intelligent recovery mechanism is introduced to ensure the normal operation of the inverter, and if the inverter cannot operate normally after the power reduction, the inverter is restored to the rated power, so that the power generation of the inverter is reduced, the inverter is prevented from being in a long-term overload state, and the power generation stability of the entire photovoltaic station area is improved. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0056] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0057] The technical solutions of the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. The components of the embodiments of the present application described and shown in the drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of the present application.

[0058] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first", "second" and the like are only used to distinguish the description, and cannot be understood as indicating or implying relative importance, or implying any such actual relationship or order between these entities or operations. In addition, the terms "connected", "connected" and the like can be direct connection between elements, or indirect connection via other elements.

[0059] Embodiment 1:

[0060] The present application is realized by the following technical solutions, such asFigure 1 As shown in the figure, an intelligent control method of inverter combining dynamic calculation of power with outlier elimination includes the following steps:

[0061] Step 1, obtain the real-time power of the inverter, and calculate the adjustable coefficient of each inverter based on the real-time power.

[0062] The real-time power of n inverters in the photovoltaic station is obtained in real time through the Modbus protocol of the inverter, and the real-time power of the i-th inverter is denoted as P new,i , wherein i=1, 2,..., n; based on the real-time power P new,i of the i-th inverter, the adjustable coefficient k i of the i-th inverter is calculated. (1)

[0063] , wherein, is the weight of the i-th inverter, , P rated,i is the rated power of the i-th inverter;

[0064] α and β are adjustable coefficient weight factors, α+β=1, and the default α=0.7 and β=0.3. Since α controls the influence weight of the real-time power of the inverter on the adjustable coefficient, if α is set too large, such as α=1 and β=0, it indicates that the adjustable coefficient of the inverter is mainly calculated through the real-time power of the inverter, which may cause the inverter with smaller rated power to bear too heavy a task and thus exceed the bearing limit of the inverter; if β is set too large, such as α=0 and β=1, it indicates that the adjustable coefficient of the inverter is mainly calculated through the rated power of the inverter, which may ignore the actual power generation capacity of the inverter and cause the inverter with smaller rated power to be not fully utilized. Therefore, the real-time light intensity is obtained by reading the inverter register address, and the values of α and β are adjusted according to the light intensity, specifically:

[0065] (1) when I<300W / m 2 , the working efficiency of the photovoltaic panel needs to be improved, and at this time α is adjusted by 0.2 based on the default value 0.7, and β is adjusted by 0.2 based on the default value 0.3, i.e. α=0.9 and β=0.1.

[0066] (2) when 300W / m 2 ≤I≤500W / m 2 , α is adjusted by 0.1 based on the default value 0.7, and β is adjusted by 0.1 based on the default value 0.3, i.e. α=0.8 and β=0.2.

[0067] (3) when 500W / m 2 <I≤800W / m2 When it is, it is the optimal light intensity. At this time, the default values are α = 0.7 and β = 0.3.

[0068] (4) When it is 800 W / m 2 <I, the power generation of the photovoltaic panel is overloaded. At this time, α is adjusted downward by 0.3 based on the default value of 0.7, and at the same time, β is adjusted upward by 0.3 based on the default value of 0.3, that is, α = 0.4 and β = 0.6.

[0069] Step 2, calculate the total power reduction amount and the total adjustable power of all inverters in the photovoltaic area, and allocate the initial reduction power to each inverter.

[0070] Subtract the upper limit value of the safe power operation of the photovoltaic area from the sum of the rated output powers of all inverters in the photovoltaic area to obtain the output overlimit value P over ; and based on the output overlimit value P over Calculate the total power reduction amount : (2) (3)

[0071] Among them, P over is the output overlimit value; P total is the rated output power of all inverters; P max is the upper limit value of the safe power operation of the photovoltaic area. When calculating the total power reduction amount , if both the output overlimit value P over and meet the conditions for reducing power, the purpose of selecting the minimum value is to reduce the waste of power generation.

[0072] Calculate the total adjustable power P adjust_total based on the maximum adjustable power of all inverters in the photovoltaic area: (4)

[0073] Among them, is the maximum adjustable power of the i-th inverter.

[0074] Based on the total power reduction amount , allocate the initial reduction power to each inverter : (5)

[0075] Store the initial reduction power allocated to all inverters into the reduction power queue U, and calculate and update the reduction power queue U every minute.

[0076] ​Step 3: Calculate the unallocated power based on the initial power reduction, and then allocate the power reduction to the inverters again based on the unallocated power to obtain the planned power reduction for each inverter.

[0077] Calculate the unallocated power P in the photovoltaic area. undistributed : (6)

[0078] in, This represents the total power reduction.

[0079] Remove real-time power P from power reduction queue U new,i For inverters with a power rating of 0, obtain the effective inverter list B, and then redistribute the power reduction to the inverters in the effective inverter list B: (7)

[0080] in, Let t be the power reduction for the i-th inverter in the effective inverter list B; since equation (5) has already performed an initial power reduction allocation, and the initial power reduction for the first allocation is shown in equation (6), then t is the power reduction for the t-th allocation. The unallocated power P obtained later undistributed Therefore, in equation (7), t≥2; P undistributed_t The unallocated power in the photovoltaic area after t power reduction allocations to the inverters in the effective inverter list B; x is the number of inverters removed from the power reduction queue U, x≥0; k i Let x be the adjustable coefficient of the i-th inverter in the effective inverter list B, where i = 1, 2, ..., n.

[0081] Construct the power loss model L of the i-th inverter i : (8)

[0082] Where a is the conversion efficiency of the i-th inverter, obtained by reading the inverter register address; b is the current electricity price; and c is the startup loss of the i-th inverter. This represents the loss caused by reducing the power of the i-th inverter, which generally decreases as the power reduction decreases. Since inverters typically experience reduced conversion efficiency when operating away from their maximum power point, a 10% power reduction might result in a 0.5% efficiency decrease, but a 50% power reduction might lead to a 3% efficiency decrease. This efficiency decline is not linear but accelerates. This simulation of nonlinear relationship is quite good. This indicates direct economic losses and power reduction. The inability to sell through the power grid results in a direct loss of revenue. c represents the start-up loss of the inverter, the power generation loss caused by the small amount of energy consumed by the inverter in running, receiving instructions, processing data, and operating control actions itself, in order to maintain the adjustable state, the inverter may need to be maintained in a ready mode, which has slightly more power consumption than the full standby mode.

[0083] Calculate the current loss L of the i-th inverter in the effective inverter list B i_now , solve the t value corresponding to the minimum value of the current loss L i_now : (9)

[0084] wherein, represents the t-th derivative of the power loss model L i . By derivation, the t-th calculated reduction power i_now corresponding to the minimum value of L is added to the initial reduction power , and the planned reduction power is updated, that is, assuming t=4, the planned reduction power is:

[0085] That is, by t times of iteration calculation of the unallocated reduction power, and allocating the unallocated reduction power to the inverter according to the constraint of the power loss model L i , the inverter can reduce the reduction loss, reduce the economic loss, and reduce the start-up loss when reducing the power.

[0086] Step 4: input the planned reduction power into the optimization model to obtain the final reduction power, and reduce the power of each inverter according to the final reduction power.

[0087] Define the objective function and two constraint conditions of the optimization model: (10) (11) (12)

[0088] Wherein, formula (10) is the objective function, formula (11) is the power balance constraint condition, and formula (12) is the power non-negative constraint condition.

[0089] The objective function Object represents the "total adjustment cost" of the system, and the reciprocal of the adjustable coefficient k i is taken as the "cost coefficient", and the calculated adjustable coefficient k i ​The greater, the greater the power that the inverter i is cut; conversely, the smaller the power that the inverter i is cut. The total regulation cost Object changes in a quadratic curve with the increase of the planned cut power. The large-scale cut power behavior of the inverters under the photovoltaic station area has a significant harm to the equipment itself and the system efficiency. On the one hand, the rough reduction instruction is easy to cause the power control to be inaccurate due to the nonlinear response of the equipment and the grid fluctuation, and even trigger the overvoltage protection and off-grid, causing the unplanned shutdown, and the long-term deep limit will also cause the internal local overheating and the capacitor failure, accelerating the equipment aging. On the other hand, the inverter has a typical "inverted U-shaped" curve efficiency, and the efficiency is optimal in the interval of 35% to 65% of the rated load, and the large-scale cut power will make it fall to the low-efficiency light load area, greatly increasing the proportion of fixed loss, causing the serious waste of light energy resources. Therefore, in the actual environment, the large-scale cut power may not get the expected value, and the large-scale cut power may damage the equipment and the efficiency conversion is lower. In order to realize scientific regulation and control, the modern photovoltaic station area system should adopt the fine cluster strategy, give priority to the call of the reactive power support to the grid, reduce the active cut, and realize the full discharge of the photovoltaic and the storage of the excess electricity, so as to improve the energy utilization efficiency while ensuring the safety of the grid.

[0090] The power balance constraint condition is to ensure that the sum of the cut power of all inverters is exactly equal to the total cut amount required, so as to realize accurate control and maintain the power quality of the grid. The power non-negative constraint condition is that the planned cut power of each inverter can only be zero or positive under the current set cut power scene.

[0091] The adjustable coefficient k i is input into the optimization model to allocate the final cut power for each inverter, and the specific calculation method is as follows:

[0092] Firstly, the Lagrange multiplier is introduced to integrate the power balance constraint condition into the objective function to form the Lagrange function: (13)

[0093] Among them, is the Lagrange function. is the marginal regulation cost.

[0094] Then, the partial derivative of the Lagrange function is taken and is set to 0: (14) (15)

[0095] Where L is the Lagrange function; the planned curtailment power of each inverter proportional to its adjustable coefficient k i proportional to its marginal regulation cost proportional to its marginal regulation cost.

[0096] In the power allocation optimization problem of inverter cluster under photovoltaic district, the Lagrange method shows its unique theoretical advantages and engineering value, which transforms the complex dynamic optimization problem with constraints into an efficient solution form with analytical solution. The first feature "constraint embedding mechanism" embeds the key physical constraint of system total power balance into the objective function in a structured way by introducing Lagrange multipliers, realizing the essential conversion from constraint optimization to unconstrained saddle point problem. The second feature "steady zero-guiding characteristics" transforms the iterative search process of traditional gradient descent into the problem of directly solving linear equations by establishing the optimization criterion that the partial derivative of Lagrange function is zero, not only reducing the problem solving dimension, but also completely avoiding the risk of numerical instability such as gradient disappearance and convergence oscillation. The third feature "analytical closed-loop solution" finally realizes the theoretical breakthrough, that is, the explicit analytical solution of power allocation is obtained directly by decoupling linear system, which has both mathematical simplicity and physical interpretability, and the calculation complexity is only linearly related to the size of inverter. The three features form a layer-by-layer progressive technical closed loop, which enables the system to complete the accurate generation of optimal power strategy within milliseconds, not only strictly following the rigid constraint of power balance of power grid, but also supporting the high-frequency demand of minute-level real-time decision with excellent numerical stability and calculation efficiency, thereby laying the foundation for the power allocation system to achieve optimality in dynamic scenarios.

[0097] Then, the partial derivative of is taken and set to 0: (16)

[0098] Substituting equation (15) into equation (16), we get: (17)

[0099] That is: (18)

[0100] Finally, substituting equation (18) into equation (15), we get: (19)

[0101] The final curtailment power of all inverters is stored in the curtailment power queue U, and the final curtailment power is calculated every minute and the curtailment power queue U is updated.

[0102] Step 5, fine-tune inverter power.

[0103] Calculate the expected power of each inverter after power reduction: (20)

[0104] Among them, P achieve,i P represents the desired power output after reduction of the i-th inverter. current,i The power of the i-th inverter before it was reduced.

[0105] After power reduction, determine the real-time power P of the inverter. new,i Is it in (P) achieve,i ×(1-t / 100), P achieve,i Within the range of P (default t=5), if it is within the range, the operation of the inverter will not be interfered with; if it is greater than P achieve,i Then step The percentage was lowered by 1%; if it is less than P achieve,i ×(1-t / 100), then the step The percentage was increased by 1%.

[0106] Step 6: Use the local outlier factor algorithm to determine whether the inverter is operating abnormally after the power reduction. If it is operating abnormally, intelligently restore the inverter's power to the rated power.

[0107] Read the real-time power P of each inverter every 5 seconds. new,i and P new,i Store the data in queue C, and calculate the average real-time power in queue C. : (twenty one)

[0108] Where N is the storage capacity of queue C, with a default value of N=10.

[0109] The MQ-LOF adaptive filtering method for filtering outlier data is adopted, and it consists of two stages:

[0110] Phase 1: Fast coarse filtering based on dynamic moving quantiles aims to quickly remove obvious and extreme outliers, reducing the computational burden of Phase 2.

[0111] In detail, maintain a recent historical data set of fixed length L (sourced from queue C), and set the sliding window to W=[P t-L ,P t-L+1 ,...,P t-1 For the data point P newly added to queue C new,i Calculate the two dynamic quantiles of window W, namely the upper quantile Q. 1-α and lower quantile Q α (Default α = 0.05). If P new,i <Qα or P new,i Q 1-α Then the data point P new,i If a data point is identified as an outlier, it is immediately removed from queue C; otherwise, the data point P is... new,i After coarse filtering, the process moves to stage two for fine identification.

[0112] Phase Two: Fine-grained identification based on the Local Outlier Factor (LOF) algorithm, which aims to identify potential outliers that are not obviously extreme but behave “strange” in their local time series context.

[0113] In detail, Phase Two is achieved through the following steps:

[0114] (1) Construct feature vectors and add the newly added data points P to queue C. new,i Together with the preceding d-1 data points, they form a d-dimensional feature vector. This feature vector characterizes the power variation pattern over a recent period. Each inverter i constructs a d-dimensional feature vector at time t. The purpose is to determine whether the state of the i-th inverter is abnormal at time t.

[0115] (2) The dataset consisting of the eigenvectors of m inverters at time t is: The Euclidean distance is used to measure the similarity of the power change patterns of any two inverters at time t: (twenty two)

[0116] in, Let be the similarity between the power change patterns of the i-th inverter and the j-th inverter at time t. Let J be the eigenvector of the j-th inverter, where j = 1, 2, ..., m; Let be the real-time power of the i-th inverter at time ts; Let be the real-time power of the j-th inverter at time ts; s = 0, 1, ..., d-1.

[0117] (3) Eigenvectors The K-distance neighborhood is It refers to all features related to the eigenvector. The distance is no greater than A set of eigenvectors, which contains at least k eigenvectors, where, express The distance to its k-th nearest neighbor.

[0118] (4) Calculate the eigenvectors Locally achievable density: (23)

[0119] wherein, is the local reachable density of the feature vector ; denotes the set of feature vectors in the K-distance neighborhood of the feature vector ; is the reachable distance from to ; is the neighbor of within the K-distance neighborhood , and .

[0120] (5) Calculate the local outlier factor (LOF) of the feature vector : (24)

[0121] wherein, is the local outlier factor of the feature vector .

[0122] If , it means that the local density of the point is comparable to the average local density of its k neighbors, that is, the power change pattern at the current t time is consistent with its k neighbors, that is, the d-dimensional real-time power in is a normal value. If , it means that the local density of the point is much lower than the average local density of its k neighbors, that is, the power change model at the current t time is very "sparse" or "isolated" relative to its k neighbors, that is, there is an abnormal value in . If , it means that the local density of the point is much higher than the average local density of the k neighbors, that is, it is in a very dense core area at the current t time, which is usually a normal value.

[0123] Preferably, if , the real-time power at t time is considered as an abnormal value, and the real-time power is removed from the queue C, wherein is a threshold value, and the default . It should be noted that the real-time power in formula (21) is recorded as "P new,i ", but in order to show the inverter to which the real-time power belongs and the collection time, P new,i is used to represent P .

[0124] After the outliers are removed from the queue C, the average of the real-time power in the queue C is recalculated based on equation (21) , if , it means that the power generated by the i-th inverter is not within the range of (P achieve,i ×(1-t / 100), P achieve,i ), thus it can be concluded that the fine-tuning mechanism cannot adjust the real-time power to be infinitely close to the expected power P achieve,i , which proves that the light intensity is in a weak period, and the power generated by the inverter has reached a low state. In order to ensure that the light resource is fully utilized the next day, the power recovery operation needs to be performed on the inverter, and the real-time power of the inverter is adjusted back to the rated power, so as to protect the safe operation of the inverter.

[0125] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which shall be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An intelligent inverter control method integrating dynamic power calculation and outlier removal, characterized in that, Includes the following steps: Step 1: Obtain the real-time power of the inverter and calculate the adjustability coefficient of each inverter based on the real-time power. Step 2: Calculate the total power reduction and total adjustable power of all inverters under the photovoltaic area, and allocate the initial power reduction to each inverter; Step 3: Calculate the unallocated power based on the initial power reduction, and then allocate the power reduction to the inverters again based on the unallocated power to obtain the planned power reduction for each inverter. Step 4: Input the planned power reduction into the optimization model to obtain the final power reduction, and then reduce the power of each inverter according to the final power reduction. Step 5: Fine-tune the inverter power; Step 6: Use the local outlier factor algorithm to determine whether the inverter is operating abnormally after the power reduction. If it is operating abnormally, intelligently restore the inverter's power to the rated power.

2. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 1, characterized in that, The formula for calculating the adjustable coefficient in step 1 is as follows: in, Let i be the weight of the i-th inverter. P rated,i Let be the rated power of the i-th inverter; n be the number of inverters in the photovoltaic area; α and β be the adjustable coefficient weighting factors.

3. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 2, characterized in that, In step 2, the total power reduction of all inverters under the photovoltaic area is calculated as follows: in, P represents the total power reduction. over To output the out-of-limit value; P total P represents the rated output power of all inverters. max This represents the upper limit of safe power operation for photovoltaic power distribution areas.

4. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 3, characterized in that, In step 2, the initial power reduction allocated to each inverter is: in, Let be the initial power reduction of the i-th inverter; Initial power reduction allocated to all inverters Stored in the power reduction queue U, calculated once per minute. And update the power reduction queue U.

5. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 4, characterized in that, In step 3, the unallocated power calculated based on the initial power reduction is: Among them, P undistributed This refers to the unallocated power in the photovoltaic distribution area.

6. The method for dynamic control and intelligent recovery of photovoltaic power overload according to claim 5, characterized in that, In step 3, the power reduction is re-allocated to the inverters based on the unallocated power, resulting in the planned power reduction for each inverter: Remove real-time power P from power reduction queue U new,i For inverters with a power rating of 0, obtain the effective inverter list B, and then redistribute the power reduction to the inverters in the effective inverter list B: in, The power reduction is the amount of power allocated to the i-th inverter in the effective inverter list B for the t-th time; t≥2; P undistributed_t The unallocated power in the photovoltaic area after t power reductions to the inverters in the effective inverter list B; x is the number of inverters removed from the power reduction queue U, x≥0; i=1,2,...,nx; Construct the power loss model L of the i-th inverter i : Where a is the conversion efficiency of the i-th inverter; b is the current electricity price; and c is the startup loss of the i-th inverter. Calculate the current loss L of the i-th inverter in the effective inverter list B. i_now Solve for the current loss L i_now Minimum value : in, This indicates the power loss model L i Find the t-th derivative; obtain L by differentiation. i_now The t value corresponding to the minimum value; The power reduction calculated t times With initial power reduction Adding them together, the updated plan yields the power reduction. .

7. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 2, characterized in that, In step 4, the planned power reduction will be implemented. The final power reduction obtained by inputting the optimization model is: Define the objective function and two constraints for the optimization model: objective function ; Power balance constraints ; Power nonnegativity constraint ; Constructing the Lagrange function: in, It is a Lagrange function; For marginal adjustment costs; Take the partial derivative of the Lagrange function and set it to 0: Where L is the Lagrange function; right Find the partial derivative and set it to 0: have: Right now: The final result is: in, This represents the final power reduction of the i-th inverter; Final power reduction allocated to all inverters Stored in the power reduction queue U, the final power reduction is calculated every minute. And update the power reduction queue U.

8. The inverter intelligent control method for integrating dynamic power calculation and outlier removal according to claim 1, characterized in that, Step 6 specifically includes the following steps: Read the inverter's real-time power P every 5 seconds. new,i and P new,i Store in queue C; Based on queue C, maintain recent historical data of fixed length L, and set the sliding window to W=[P t-L ,P t-L+1 ,...,P t-1 For the data point P newly added to queue C new,i Calculate the two dynamic quantiles of window W, namely the upper quantile Q. 1-α and lower quantile Q α If P new,i α or P new,i Q 1-α Then the data point P new,i It was identified as an outlier and the data point was directly removed from queue C;​ Otherwise, the local outlier factor algorithm is used to identify whether the data point is abnormal. If it is abnormal, the data point is removed from queue C. After removing outliers from queue C, calculate the average real-time power in queue C. : Where N is the storage capacity of queue C; like Adjust the real-time power of the inverter back to its rated power, where P achieve,i For the desired power output of the i-th inverter after reduction, we have: P current,i The power of the i-th inverter before it was reduced is given.

Images