A two-layer automatic power generation control method for photovoltaic power stations based on data-driven voltage feedback

Abstract

A two-layer automatic power generation control method for photovoltaic power stations based on data-driven voltage feedback is a method for calculating the linear power flow matrix of photovoltaic power stations based on linear dimensionality-increasing transformation of state space. Based on the power flow matrix calculation results and the acquisition of the photovoltaic power station's operating status, a photovoltaic array active power optimization control model is constructed in the photovoltaic power station's automatic power generation control system. The model is solved using the interior point method. At the same time, the voltage-active power sensitivity is calculated based on the linear power flow matrix. The automatic power generation control system transmits the active power, grid-connected point voltage amplitude, and the sensitivity of voltage amplitude to active power of each photovoltaic array to each photovoltaic array via a communication channel. At the photovoltaic array controller level, based on the deviation between the actual measured voltage amplitude and the voltage amplitude transmitted by the automatic power generation control system, the active power secondary adjustment amount is calculated based on the received sensitivity and transmitted to each photovoltaic inverter. The present invention can quickly correct control deviations without the need for the system to re-optimize the calculation.

Description

Technical Field

[0001] The present invention relates to a double-layer automatic power generation control method for a photovoltaic power station, and more particularly to a double-layer automatic power generation control method for a photovoltaic power station based on data-driven voltage feedback. Background Art

[0002] With increasing attention to environmental pollution and the comprehensive advancement of smart grid construction, the installed capacity and grid-connected power generation of renewable energy generation have continued to increase. The time-varying and complex nature of renewable energy generation operations has become increasingly prominent, significantly increasing the operational risks and control difficulties of the power grid. Among these, the mismatch between unstable primary energy input and grid-connection assessment requirements is currently a key issue hindering the high-quality grid connection of large-scale photovoltaic power plants. The "Implementation Rules for the Management of Grid-Connected Operation of Power Plants" and "Implementation Rules for the Management of Auxiliary Services for Grid-Connected Power Plants" of each regional power grid have set specific requirements for photovoltaic power plants to participate in Automatic Generation Control (AGC).

[0003] Due to the large number of PV inverters in large-scale PV power plants, their high volatility and susceptibility to grid disconnection, active power control in these plants presents systemic challenges. The grid-connected power quality requirements of PV power plants necessitate continuous improvements in the functional completeness and technological sophistication of automatic power generation control systems. This is particularly true for PV power plants operating under relatively complex climatic conditions, where the wide fluctuation range of primary energy and large instantaneous power variations make it prone to large power fluctuations during plant operation. This not only makes it difficult to respond to AGC commands, but also makes it prone to line overloads or grid connection point overvoltage issues within the PV plant, increasing the likelihood of PV array disconnection and posing operational risks to the PV plant. Therefore, increasingly complex system structures and increasingly stringent power quality requirements have necessitated the urgent need for technological innovation in automatic voltage control methods to ensure plant response to dispatching commands and stable, economical operation.

[0004] Currently, automatic power generation control systems for photovoltaic power plants have been developed and deployed on a large scale, and traditional technologies are relatively mature. However, existing automatic power generation control algorithms generally allocate power proportionally to the capacity of the photovoltaic arrays based on AGC commands issued by the dispatcher. This fails to consider the actual operating conditions of each photovoltaic array. As a result, the issued control commands are prone to failure to execute, or the resulting execution results in line overload or voltage limit violations. In the field of technical research, model-based automatic power generation control methods have emerged in large numbers, but most fail to consider the actual communication constraints of photovoltaic power plants and the accuracy of model parameters. If the issued active power adjustment command fails to execute due to packet loss due to communication problems, or if the execution result is unsatisfactory due to inaccurate model parameters, the effectiveness of the photovoltaic power plant's automatic power generation control can be compromised, even posing safety risks. Due to the operating mechanism and communication method of automatic power generation control systems, their command cycles are generally longer than the speed of operating state changes, making it difficult to compensate for these issues through re-optimization.

[0005] To address these issues, driven by the need for technological innovation in automatic power generation control and constrained by practical communication and parameter requirements, a practical two-tier control approach is urgently needed. This approach uses the PV power plant's automatic power generation control system to issue active power adjustment commands and set voltage values ​​for the PV array's grid connection point. Each PV array can then rapidly implement secondary power adjustments through voltage deviation feedback, thereby achieving automatic power generation control that takes into account both the external characteristics and internal distribution of the PV power plant. Furthermore, to address the issue of imprecise model parameters, a data-driven state-space transformation model identification method can be used to accurately calculate the sensitivity relationship between the PV array's grid connection point voltage and active power. This approach can then assist in implementing secondary active power adjustments based on voltage feedback, ultimately achieving greater engineering applicability. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a photovoltaic power station double-layer automatic power generation control method based on data-driven voltage feedback, which can quickly correct control result deviations caused by communication problems and inaccurate parameters in order to overcome the shortcomings of the existing technology.

[0007] The technical solution adopted by the present invention is: a double-layer automatic power generation control method for a photovoltaic power station based on data-driven voltage feedback, comprising:

[0008] 1) Based on the selected PV power station, the PV array access location and capacity, the PV array capacity and the upper limit of each PV inverter capacity within the PV array, the upper and lower voltage limits of the PV array grid connection point, the PV power station collector line impedance parameters, the upper limit of the PV power station collector line transmission capacity, and the rated voltage of the PV array grid connection point are input into the PV power station's automatic power generation control system. In addition, the PV power station's automatic power generation control system obtains in real time the AGC instructions issued by the dispatch center to the automatic power generation control system, the maximum active power value of each PV array, the measured voltage value of each PV array grid connection point, and the reactive power of each PV array.

[0009] 2) The automatic power generation control system of the photovoltaic power station uses the historical measurement data of the voltage phase angle and amplitude at the photovoltaic array grid connection point and the active and reactive power of the photovoltaic array as training samples. The input variables of the training samples are transformed by the Koopman state space dimensionality increase method, and the linear power flow equation matrix is ​​constructed based on the least squares method.

[0010] 3) The automatic power generation control system of the photovoltaic power station builds a photovoltaic array active power optimization control model based on the AGC instructions issued by the dispatching center and the maximum active power value of each photovoltaic array, including: objective function and constraints, and uses the interior point method to solve the photovoltaic array active power optimization control model to obtain the active power of each photovoltaic array and the voltage amplitude V of each photovoltaic array grid connection point. i ;

[0011] 4) Based on the results of the PV array active power optimization control model and the value of the linear power flow equation matrix L, the real-time sensitivity of the voltage at each PV array grid connection point to the PV array active power is calculated:

[0012]

[0013] Among them, Z ii is the voltage amplitude V at the grid connection point of the i-th photovoltaic array i For the active power p of the i-th photovoltaic array i Sensitivity, L ii is the voltage amplitude V corresponding to the grid connection point of the i-th photovoltaic array in the linear power flow equation matrix L i and the active power p of the i-th photovoltaic array i The element, c tv is a randomly generated K-dimensional basis vector c t The vth element in c ti is a randomly generated K-dimensional basis vector c t The i-th element in is the input variable value of the current time section, The input variable value for the current time section The vth element in, K is the dimension of the input variable of the current time section, L i,(K+t) is the voltage amplitude V corresponding to the grid connection point of the ith photovoltaic array in the matrix linear power flow equation matrix L i and the t-th dimension input variable ψ expanded by the dimensionality-raising transformation t (x) elements, m is the dimension of the ascending dimension;

[0014] 5) The automatic power generation control system of the photovoltaic power station converts the active power p of each photovoltaic array into i , the voltage amplitude of each photovoltaic array grid connection point V i , and the sensitivity of the voltage amplitude of each photovoltaic array grid-connected point to the active power of the photovoltaic array is sent to each photovoltaic array controller through the photovoltaic power station communication channel;

[0015] 6) Each photovoltaic array controller receives the corresponding active power p i As a result, the active power of each photovoltaic inverter in the corresponding photovoltaic array is adjusted according to the following formula:

[0016]

[0017] Among them, p ij is the active power of the jth PV inverter in the i-th PV array, S ij,max is the maximum capacity of the jth PV inverter in the i-th PV array, S i,max is the maximum capacity of the i-th photovoltaic array;

[0018] 7) Each photovoltaic array controller calculates the active power adjustment of the corresponding photovoltaic array based on the deviation between the actual value of the voltage amplitude at the corresponding photovoltaic array grid connection point and the voltage command issued by the photovoltaic power station automatic power generation control system, and performs secondary adjustment on the active power of each photovoltaic inverter in the photovoltaic array according to the active power adjustment;

[0019] 8) Each photovoltaic array controller adjusts the active power of the photovoltaic inverter according to the secondary adjustment amount of the active power of the photovoltaic inverter, and returns to step 7).

[0020] The present invention's dual-layer automatic power generation control method for photovoltaic power stations based on data-driven voltage feedback can quickly correct control result deviations caused by communication problems and parameter inaccuracies without requiring the system to re-optimize calculations. Furthermore, the data-driven power flow matrix used is independent of model parameter accuracy, resulting in greater engineering applicability. The advantages and benefits of the present invention are:

[0021] 1. Based on the traditional automatic power generation control of large-scale photovoltaic power stations, the present invention takes into account the optimal distribution of photovoltaic array power within the photovoltaic power station, so that the photovoltaic power station as a whole can meet the AGC instructions issued by the dispatching center while achieving optimized operation within the photovoltaic power station. In particular, it can take into account the voltage constraints of the grid connection point and the capacity constraints of the collection line while performing automatic power generation control.

[0022] 2. The present invention proposes a two-layer control architecture for automatic power generation control based on grid-connected point voltage feedback. The upper layer is implemented by the automatic power generation control system of the photovoltaic power station, and the lower layer is implemented by the photovoltaic array controller. It fully considers the situation where the active power instructions issued by the automatic power generation control system are lost due to actual communication constraints or cannot be executed due to restrictions on the operating conditions of the photovoltaic inverter. The photovoltaic array controller detects the local grid-connected point voltage and uses voltage feedback to achieve secondary adjustment of active power. Since there is no need for the automatic power generation control system to re-centrally collect and optimize, the secondary adjustment speed is significantly improved, and through periodic cycles, it can ultimately achieve zero-difference adjustment between actual control and theoretical optimization results under communication constraints.

[0023] 3. The data-driven grid-connected point voltage and photovoltaic array active power sensitivity calculation method proposed in the present invention fully utilizes the historical operating data samples of the photovoltaic power station and constructs a dimensional linearized photovoltaic power station power flow model. The sensitivity calculation results are not only valid for a certain operating point, but also achieve global high-precision applicability. The Koopman data-driven method adopted does not depend on the model parameters. When there are errors in the actual photovoltaic power station collection line parameters, the accuracy of this method is significantly improved compared with the traditional sensitivity analysis method. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 This is a flow chart of a double-layer automatic power generation control method for a photovoltaic power station based on data-driven voltage feedback according to the present invention. DETAILED DESCRIPTION

[0025] The following describes in detail the double-layer automatic power generation control method of a photovoltaic power station based on data-driven voltage feedback according to the present invention in conjunction with the embodiments and drawings.

[0026] The present invention proposes a two-tiered automatic power generation control method for photovoltaic power plants based on data-driven voltage feedback. This method proposes a linear power flow matrix calculation method for photovoltaic power plants based on a linear dimensionality-increasing transformation of state space. Based on the power flow matrix calculation results and the collected operational status of the photovoltaic power plant, a photovoltaic array active power optimization control model is constructed within the photovoltaic power plant's automatic power generation control system. The model is solved using the interior point method. Simultaneously, the voltage-active power sensitivity is calculated based on the linear power flow matrix. The automatic power generation control system transmits the active power, grid-connection point voltage amplitude, and the voltage amplitude-to-active power sensitivity of each photovoltaic array via a communication channel. At the photovoltaic array controller level, based on the deviation between the actual measured voltage amplitude and the voltage amplitude transmitted by the automatic power generation control system, a secondary active power adjustment is calculated based on the received sensitivity and transmitted to each photovoltaic inverter. This process can be repeated based on voltage variations. Compared to methods based solely on automatic power generation control systems, this method can quickly correct control result deviations caused by communication issues and parameter inaccuracies, eliminating the need for system recalculation. Furthermore, the data-driven power flow matrix employed is independent of model parameter accuracy, resulting in greater engineering applicability.

[0027] like Figure 1 As shown, the double-layer automatic power generation control method of a photovoltaic power station based on data-driven voltage feedback of the present invention includes the following steps:

[0028] 1) Based on the selected PV power station, the PV array access location and capacity, the PV array capacity and the upper limit of each PV inverter capacity within the PV array, the upper and lower voltage limits of the PV array grid connection point, the PV power station collector line impedance parameters, the upper limit of the PV power station collector line transmission capacity, and the rated voltage of the PV array grid connection point are input into the PV power station's automatic power generation control system. In addition, the PV power station's automatic power generation control system obtains in real time the AGC instructions (automatic power generation control instructions) issued by the dispatch center to the automatic power generation control system, the maximum active power value of each PV array, the measured voltage value of each PV array grid connection point, and the reactive power of each PV array.

[0029] 2) The automatic power generation control system of the photovoltaic power station uses the historical measurement data of the voltage phase angle and amplitude at the photovoltaic array grid connection point and the active and reactive power of the photovoltaic array as training samples. The input variables of the training samples are transformed by the Koopman state space dimensionality increase method, and the linear power flow equation matrix is ​​constructed based on the least squares method.

[0030] (1) The formula for performing dimensionality-increasing transformation on the input variables of the training samples is as follows:

[0031]

[0032] Where x is the input variable of a time section, including the historical measurement data of active and reactive power of the photovoltaic array, ψ(x) represents the m-dimensional input variable expanded by dimensionality increase, and x lift is the input variable after dimensionality increase transformation of a time section;

[0033] The t-th dimension input variable expanded by the dimensionality-raising transformation is calculated as follows:

[0034]

[0035] Where, ψ t (x) is the t-th element in the m-dimensional input variable ψ(x) obtained by dimensional expansion, t=1,2,3,...m, f lift is a dimensionality-raising function, x v is the vth element in the input variable x of a time section, K is the dimension of the input variable x of a time section, c t is a randomly generated K-dimensional basis vector, c tv is a randomly generated K-dimensional basis vector c t The vth element in .

[0036] (2) The linear power flow equation matrix is ​​constructed based on the least squares method, as shown in the following formula:

[0037]

[0038] Where L is the linear power flow equation matrix, is the input variable sample matrix, which contains the active and reactive power vectors of the photovoltaic array at S time sections. is the input variable sample matrix after dimensionality increase, which contains the input variables after dimensionality increase of S time sections, where S is the number of time sections. is the state variable sample matrix, which contains the voltage phase angle and amplitude vector of the photovoltaic array grid connection point at S time sections, [·] + is the Moore-Penrose inverse of the matrix.

[0039] 3) The automatic power generation control system of the photovoltaic power station builds a photovoltaic array active power optimization control model based on the AGC instructions issued by the dispatching center and the maximum active power value of each photovoltaic array, including: objective function and constraints, and uses the interior point method to solve the photovoltaic array active power optimization control model to obtain the active power of each photovoltaic array and the voltage amplitude V of each photovoltaic array grid connection point. i ;in,

[0040] (1) The objective function is as follows:

[0041]

[0042] Among them, p i is the active power of the ith photovoltaic array, and p is the power of p i The column vector is composed of N, the number of photovoltaic arrays in the photovoltaic power station, and p AGC It is the AGC instruction issued by the dispatch center, V i is the voltage amplitude of the grid-connected point of the ith PV array, μ is the rated voltage of the grid-connected point of the PV array, and ε is the voltage deviation penalty coefficient;

[0043] (2) The constraints mentioned include:

[0044] (2.1) Constraints on the voltage-power equation at the photovoltaic array grid connection point:

[0045]

[0046] Among them, θ is the phase angle column vector of each photovoltaic array grid connection point, V ... i The column vector is composed of q, which is the reactive power column vector of each photovoltaic array; p is the active power p of the i-th photovoltaic array. i The column vector formed, ψ(p,q) is the column vector expanded by the dimension increase of p and q;

[0047] (2.2) Constraints on the power equation of the photovoltaic power station collection line:

[0048]

[0049]

[0050] Among them, P ij is the active power transmitted by the collector line between the i-th photovoltaic array and the j-th photovoltaic array, Q ij is the reactive power transmitted by the collector line between the i-th photovoltaic array and the j-th photovoltaic array, q j is the reactive power of the jth photovoltaic array, r ij and x ij are the resistance and reactance of the collector line between the i-th photovoltaic array and the j-th photovoltaic array respectively;

[0051] (2.3) Voltage constraints at the photovoltaic array grid connection point:

[0052] V min ≤V i ≤V max ,i∈{1,2,…,N} (8)

[0053] Among them, V min and V max are the upper and lower voltage limits of the photovoltaic array grid connection point, V i is the voltage amplitude of the grid-connected point of the i-th photovoltaic array;

[0054] (2.4) Capacity constraints of photovoltaic power station collection lines:

[0055]

[0056] Where N is the number of photovoltaic arrays in the photovoltaic power station, S max The upper limit of the capacity of the photovoltaic power station collection line;

[0057] (2.5) PV array active power constraint:

[0058] 0≤p i ≤p i,max ,i∈{1,2,…,N} (10) Among them, p i is the active power of the ith photovoltaic array, p i,max is the maximum active power that can be generated by the i-th photovoltaic array;

[0059] (2.6) PV array capacity constraints:

[0060]

[0061] Among them, q i is the reactive power of the ith photovoltaic array, S i,max is the maximum capacity of the i-th photovoltaic array;

[0062] (3) The automatic power generation control system of the photovoltaic power station uses the interior point method to solve the photovoltaic array active power optimization control model (4) to (11), and the active power p of each photovoltaic array is obtained. i , the voltage amplitude of each photovoltaic array grid connection point V i .

[0063] 4) Based on the results of the PV array active power optimization control model and the value of the linear power flow equation matrix L, the real-time sensitivity of the voltage at each PV array grid connection point to the PV array active power is calculated:

[0064]

[0065] Among them, Z ii is the voltage amplitude V at the grid connection point of the i-th photovoltaic array i For the active power p of the i-th photovoltaic array i Sensitivity, L ii is the voltage amplitude V corresponding to the grid connection point of the i-th photovoltaic array in the linear power flow equation matrix L i and the active power p of the i-th photovoltaic array i The element, c tv is a randomly generated K-dimensional basis vector c t The vth element in c tiis a randomly generated K-dimensional basis vector c t The i-th element in is the input variable value of the current time section, The input variable value for the current time section The vth element in, K is the dimension of the input variable of the current time section, L i,(K+t) is the voltage amplitude V corresponding to the grid connection point of the ith photovoltaic array in the matrix linear power flow equation matrix L i and the t-th dimension input variable ψ expanded by the dimensionality-raising transformation t (x) elements, m is the dimension of the ascending dimension;

[0066] 5) The automatic power generation control system of the photovoltaic power station converts the active power p of each photovoltaic array into i , the voltage amplitude of each photovoltaic array grid connection point V i , and the sensitivity of the voltage amplitude of each photovoltaic array grid-connected point to the active power of the photovoltaic array is sent to each photovoltaic array controller through the photovoltaic power station communication channel;

[0067] 6) Each photovoltaic array controller receives the corresponding active power p i As a result, the active power of each photovoltaic inverter in the corresponding photovoltaic array is adjusted according to the following formula:

[0068]

[0069] Among them, p ij is the active power of the jth PV inverter in the i-th PV array, S ij,max is the maximum capacity of the jth PV inverter in the i-th PV array, S i,max is the maximum capacity of the i-th photovoltaic array;

[0070] 7) Each photovoltaic array controller calculates the active power adjustment of the corresponding photovoltaic array based on the deviation between the actual value of the voltage amplitude at the corresponding photovoltaic array grid connection point and the voltage command issued by the photovoltaic power station automatic power generation control system, and performs secondary adjustment on the active power of each photovoltaic inverter in the photovoltaic array according to the active power adjustment;

[0071] (1) The calculation formula for the active power adjustment of the photovoltaic array is as follows:

[0072]

[0073] Where Δp i is the active power adjustment of the ith photovoltaic array, V i is the voltage amplitude of the grid-connected point of the i-th photovoltaic array, is the actual value of the voltage amplitude at the grid-connected point of the i-th photovoltaic array, Z iiis the voltage amplitude V at the grid connection point of the i-th photovoltaic array i For the active power p of the i-th photovoltaic array i Sensitivity,

[0074] (2) According to the active power adjustment amount, the active power of each photovoltaic inverter in the photovoltaic array is adjusted twice, and the formula is as follows:

[0075]

[0076] Where Δp ij is the secondary adjustment of the active power of the jth photovoltaic inverter in the i-th photovoltaic array, S ij,max is the maximum capacity of the jth PV inverter in the i-th PV array, S i,max is the maximum capacity of the i-th photovoltaic array, Δp i is the active power adjustment of the i-th PV array.

[0077] 8) Each photovoltaic array controller adjusts the active power of the photovoltaic inverter according to the secondary adjustment amount of the active power of the photovoltaic inverter, and returns to step 7).

Claims

1. A dual-layer automatic power generation control method for a photovoltaic power station based on data-driven voltage feedback, characterized in that: The steps include: 1) Based on the selected PV power station, the PV array access location and capacity, the PV array capacity and the upper limit of each PV inverter capacity within the PV array, the upper and lower voltage limits of the PV array grid connection point, the PV power station collector line impedance parameters, the upper limit of the PV power station collector line transmission capacity, and the rated voltage of the PV array grid connection point are input into the PV power station's automatic power generation control system. The PV power station's automatic power generation control system then obtains in real time the AGC instructions issued by the dispatch center to the automatic power generation control system, the maximum active power value of each PV array, the measured voltage value of each PV array grid connection point, and the reactive power of each PV array. 2) The automatic power generation control system of the photovoltaic power station uses the historical measurement data of the voltage phase angle and amplitude at the photovoltaic array grid connection point and the active and reactive power of the photovoltaic array as training samples. The input variables of the training samples are transformed by the Koopman state space dimensionality increase method, and the linear power flow equation matrix is ​​constructed based on the least squares method. 3) The automatic power generation control system of the photovoltaic power station builds a photovoltaic array active power optimization control model based on the AGC instructions issued by the dispatching center and the maximum active power value of each photovoltaic array, including: objective function and constraints, and uses the interior point method to solve the photovoltaic array active power optimization control model to obtain the active power of each photovoltaic array and the voltage amplitude V of each photovoltaic array grid connection point. i ; 4) Based on the results of the PV array active power optimization control model and the value of the linear power flow equation matrix L, the real-time sensitivity of the voltage at each PV array grid connection point to the PV array active power is calculated: Among them, Z ii is the voltage amplitude V at the grid connection point of the i-th photovoltaic array i For the active power p of the i-th photovoltaic array i Sensitivity, L ii is the voltage amplitude V corresponding to the grid connection point of the i-th photovoltaic array in the linear power flow equation matrix L i and the active power p of the i-th photovoltaic array i The element, c tv is a randomly generated K-dimensional basis vector c t The vth element in c ti is a randomly generated K-dimensional basis vector c t The i-th element in is the input variable value of the current time section, The input variable value for the current time section The vth element in, K is the dimension of the input variable of the current time section, L i,(K+t) is the voltage amplitude V corresponding to the grid connection point of the ith photovoltaic array in the matrix linear power flow equation matrix L i and the t-th dimension input variable ψ expanded by the dimensionality-raising transformation t (x) elements, m is the dimension of the ascending dimension; 5) The automatic power generation control system of the photovoltaic power station converts the active power p of each photovoltaic array into i , the voltage amplitude of each photovoltaic array grid connection point V i , and the sensitivity of the voltage amplitude of each photovoltaic array grid-connected point to the active power of the photovoltaic array is sent to each photovoltaic array controller through the photovoltaic power station communication channel; 6) Each photovoltaic array controller receives the corresponding active power p i As a result, the active power of each photovoltaic inverter in the corresponding photovoltaic array is adjusted according to the following formula: Among them, p ij is the active power of the jth PV inverter in the i-th PV array, S ij,max is the maximum capacity of the jth PV inverter in the i-th PV array, S i,max is the maximum capacity of the i-th photovoltaic array; 7) Each photovoltaic array controller calculates the active power adjustment of the corresponding photovoltaic array based on the deviation between the actual value of the voltage amplitude at the corresponding photovoltaic array grid connection point and the voltage command issued by the photovoltaic power station automatic power generation control system, and performs secondary adjustment on the active power of each photovoltaic inverter in the photovoltaic array according to the active power adjustment; 8) Each photovoltaic array controller adjusts the active power of the photovoltaic inverter according to the secondary adjustment amount of the active power of the photovoltaic inverter, and returns to step 7).

2. The photovoltaic power station double-layer automatic power generation control method based on data-driven voltage feedback according to claim 1 is characterized in that: The formula for performing dimensionality-increasing transformation on the input variables of the training samples in step 2) is as follows: Where x is the input variable of a time section, including the historical measurement data of active and reactive power of the photovoltaic array, ψ(x) represents the m-dimensional input variable expanded by dimensionality increase, and x lift is the input variable after dimensionality increase transformation of a time section; The t-th dimension input variable expanded by the dimensionality-raising transformation is calculated as follows: Where, ψ t (x) is the t-th element in the m-dimensional input variable ψ(x) obtained by dimensional expansion, t=1,2,3,...m, f lift is a dimensionality-raising function, x v is the vth element in the input variable x of a time section, K is the dimension of the input variable x of a time section, c t is a randomly generated K-dimensional basis vector, c tv is a randomly generated K-dimensional basis vector c t The vth element in .

3. The photovoltaic power station double-layer automatic power generation control method based on data-driven voltage feedback according to claim 2 is characterized in that: The linear power flow equation matrix is ​​constructed based on the least squares method in step 2), as shown below: Where L is the linear power flow equation matrix, X is the input variable sample matrix, which contains the active and reactive power vectors of the photovoltaic array at S time sections, and X lift is the input variable sample matrix after dimensionality increase, which contains the input variables after dimensionality increase of S time sections, S is the number of time sections, Y is the state variable sample matrix, which contains the voltage phase angle and amplitude vector of the photovoltaic array grid connection point at S time sections, [·] + is the Moore-Penrose inverse of the matrix.

4. The photovoltaic power station double-layer automatic power generation control method based on data-driven voltage feedback according to claim 3 is characterized in that: The objective function in step 3) is as follows: Among them, p i is the active power of the ith photovoltaic array, and p is the power of p i The column vector is composed of N, the number of photovoltaic arrays in the photovoltaic power station, and p AGC It is the AGC instruction issued by the dispatch center, V i is the voltage amplitude of the grid-connected point of the ith PV array, μ is the rated voltage of the grid-connected point of the PV array, and ε is the voltage deviation penalty coefficient.

5. The double-layer automatic power generation control method for photovoltaic power station based on data-driven voltage feedback according to claim 4 is characterized in that: The constraints described in step 3) include: (1) Voltage-power equation constraints at the photovoltaic array grid connection point: Among them, θ is the phase angle column vector of each photovoltaic array grid connection point, V ... i The column vector is composed of q, which is the reactive power column vector of each photovoltaic array; p is the active power p of the i-th photovoltaic array. i The column vector is formed, ψ(p,q) is the column vector expanded by the dimension increase of p and q; L is the linear power flow equation matrix; (2) Constraints on the power equation of the photovoltaic power station collector line: Among them, P ij is the active power transmitted by the collector line between the i-th photovoltaic array and the j-th photovoltaic array, Q ij is the reactive power transmitted by the collector line between the i-th photovoltaic array and the j-th photovoltaic array, q j is the reactive power of the jth photovoltaic array, r ij and x ij are the resistance and reactance of the collector line between the i-th photovoltaic array and the j-th photovoltaic array respectively; (3) Voltage constraints at the photovoltaic array grid connection point: V min ≤V i ≤V max ,i∈{1,2,…,N} (8) Among them, V min and V max are the upper and lower voltage limits of the photovoltaic array grid connection point, V i is the voltage amplitude of the grid-connected point of the i-th photovoltaic array; (4) Capacity constraints of photovoltaic power station collection lines: Where N is the number of photovoltaic arrays in the photovoltaic power station, S max The upper limit of the capacity of the photovoltaic power station collection line; (5) PV array active power constraints: 0≤p i ≤p i,max ,i∈{1,2,…,N} (10) Among them, p i is the active power of the ith photovoltaic array, p i,max is the maximum active power that can be generated by the i-th photovoltaic array; (6) PV array capacity constraints: Among them, q i is the reactive power of the ith photovoltaic array, S i,max is the maximum capacity of the i-th photovoltaic array.

6. The photovoltaic power station double-layer automatic power generation control method based on data-driven voltage feedback according to claim 5 is characterized in that: In step 7) (1) The calculation formula for the active power adjustment of the photovoltaic array is as follows: Where Δp i is the active power adjustment of the ith photovoltaic array, V i is the voltage amplitude of the grid-connected point of the i-th photovoltaic array, is the actual value of the voltage amplitude at the grid-connected point of the i-th photovoltaic array, Z ii is the voltage amplitude V at the grid connection point of the i-th photovoltaic array i For the active power p of the i-th photovoltaic array i Sensitivity, (2) According to the active power adjustment amount, the active power of each photovoltaic inverter in the photovoltaic array is adjusted twice, and the formula is as follows: Where Δp ij is the secondary adjustment of the active power of the jth photovoltaic inverter in the i-th photovoltaic array, S ij,max is the maximum capacity of the jth PV inverter in the i-th PV array, S i,max is the maximum capacity of the i-th photovoltaic array, Δp i is the active power adjustment of the i-th PV array.

Images