Photovoltaic output prediction method and distributed optical storage transformer area active / reactive support capability optimization method

Abstract

The invention provides a photovoltaic output prediction method and a distributed optical storage transformer area active / reactive support capability optimization method, and belongs to the technical field of power system automation. Comprising the following steps: constructing a prediction model based on meteorological information, extracting multi-scale features of meteorological data, reducing dimensions through kernel principal component analysis, predicting photovoltaic power through a long-short-term memory network, quantifying the uncertainty of the photovoltaic power, and forming a prediction interval. Based on a photovoltaic power prediction result, a robust nonlinear programming model is established, the sum of absolute values of active / reactive power transmitted to a main power grid by a light storage area is maximized as a target, and linear and nonlinear constraints such as energy storage operation, power balance, equipment capacity and power factors are considered at the same time. The robustness under the photovoltaic output uncertainty is ensured through a budget uncertainty set; the model is solved by adopting a sequential quadratic programming algorithm. According to the method, the photovoltaic power prediction precision is effectively improved, and organic combination of high-precision prediction and robust nonlinear optimization is realized.

Description

Technical Field

[0001] This invention relates to a photovoltaic power output prediction method and a method for optimizing the active / reactive power support capacity of distributed photovoltaic-storage power distribution areas, belonging to the field of power system automation technology. Background Technology

[0002] Against the backdrop of the nation's vigorous promotion of energy structure transformation, the scale of distributed renewable energy, represented by photovoltaic (PV) power generation, is growing rapidly. Distributed PV-storage (PV-storage) distribution areas, as key units integrating distributed power sources and energy storage, are playing an increasingly important role in improving renewable energy absorption and enhancing the flexibility of the distribution network. However, the inherent intermittency and volatility of distributed PV power generation pose significant challenges to the distribution network, such as voltage exceeding limits and increased power fluctuations, affecting the safe and stable operation of the grid. Therefore, researching how to improve the active / reactive power support capacity of distributed PV-storage distribution areas to the grid, in order to mitigate PV fluctuations and stabilize grid operation, is of great significance.

[0003] The key to achieving effective support capabilities for distributed photovoltaic and energy storage systems lies in two aspects: first, accurately predicting photovoltaic output; and second, optimizing the operation strategy of the energy storage system to maximize the support effect.

[0004] In photovoltaic (PV) power output prediction, existing research has evolved from traditional statistical methods to approaches based on machine learning and deep learning. While deep learning models have shown advantages in capturing nonlinear time-series features, effectively extracting feature information at different scales, reducing data redundancy, and accurately quantifying prediction uncertainties remain technical challenges in the face of complex and variable weather conditions and the inherent volatility of PV power. Directly inputting raw high-dimensional data into the model may lead to low training efficiency or overfitting, affecting the robustness of the prediction.

[0005] In optimizing the active / reactive power support capacity of photovoltaic (PV) and energy storage (ESS) power grid distribution areas, existing research typically models this as an optimization problem. To address PV uncertainties, stochastic optimization or robust optimization methods have been introduced. However, considering the operational constraints of the energy storage system and inverters, as well as power balance relationships, optimization models often contain nonlinear constraints. Integrating PV output uncertainty information into robust optimization models of such nonlinear programming, while ensuring the robustness of the optimization results in the worst-case scenario without being overly conservative, and employing efficient algorithms to solve the problem, to maximize support capacity in real-time or near-real-time scheduling, presents a significant challenge for current research. Existing research rarely systematically combines high-precision prediction results with uncertainty quantification with optimization models that consider nonlinearity and robustness to directly improve the active / reactive power support capacity of PV / ESS power grid distribution areas.

[0006] Based on the above analysis, the problems and shortcomings of the existing technology are as follows: (1) Existing photovoltaic power prediction methods are insufficient in combining a large number of meteorological conditions and effectively quantifying prediction uncertainties, making it difficult to provide reliable uncertainty information to support subsequent optimization.

[0007] (2) Research on the optimization of the active / reactive power support capability of distributed photovoltaic and energy storage areas has not been able to fully and effectively integrate the uncertainty information of photovoltaic prediction with the optimization model containing nonlinear constraints. There is a lack of research on systematically combining high-precision prediction results with uncertainty quantification with optimization models that consider nonlinearity and robustness in order to directly improve the active / reactive power support capability of photovoltaic and energy storage areas to the power grid. Summary of the Invention

[0008] The purpose of this invention is to provide a photovoltaic power output prediction method and an optimization method for the active / reactive power support capability of distributed photovoltaic and energy storage areas, which can significantly improve the active / reactive power support capability of distributed photovoltaic and energy storage areas to the power grid, and enhance the stability, flexibility and renewable energy absorption capacity of the power grid.

[0009] To achieve the above objectives, the present invention employs the following technical solution: A method for predicting photovoltaic power output includes the following steps: Historical photovoltaic power data and various environmental meteorological parameter data are cleaned, outliers and missing values ​​are processed, and a training dataset is formed. The environmental meteorological parameter data are subjected to EMD decomposition to generate a feature sequence composed of intrinsic mode functions and residual terms; The feature sequence is reduced in dimensionality using a multinomial kernel function while retaining principal components; The dimensionality-reduced data is input into a long short-term memory network for training, a time series prediction model is constructed, and the photovoltaic power prediction value is output. The uncertainty of the prediction results is quantified, and the confidence interval of the prediction value is generated based on the calculated Z-score and standard deviation of the prediction value.

[0010] Preferably, the environmental meteorological parameters include: air temperature, azimuth angle, cloud opacity, dew point temperature, solar diffuse radiation index, solar direct radiation index, total horizontal solar radiation, fixed tilt radiation, tracking tilt radiation, atmospheric precipitable water, relative humidity, snow depth, surface air pressure, wind direction at 10m altitude, wind speed at 10m altitude, and zenith angle.

[0011] Preferably, the specific process of EMD decomposition includes: Locate the local maxima and minima of the input signal; The upper and lower envelopes are generated using spline interpolation. The mean line of the signal is obtained by averaging the upper and lower envelopes. : , in, Indicates the upper envelope. Indicates the lower envelope; Subtract from the original signal Extract the detail components and iterate to obtain the IMF components and residual terms; The output is a 158-dimensional feature sequence consisting of 142-dimensional IMF components and 16-dimensional residual terms.

[0012] Preferably, the KPCA uses a polynomial kernel function of order d=18, resulting in a data dimension of 26 after dimensionality reduction.

[0013] Preferably, the LSTM model network structure parameters are as follows: input layer time step 5, input dimension 9, single hidden layer, number of hidden layer units 100, output dimension 1; The Adam optimizer is used to update the weights, with a batch size of 25 and 100 training epochs.

[0014] A method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area includes the following steps: The photovoltaic power prediction value is obtained based on the photovoltaic output prediction method, and the uncertainty of the prediction result is quantified. The objective function is set to maximize the sum of the absolute values ​​of active power and reactive power of the interaction between the distributed photovoltaic energy storage area and the main power grid within the scheduling cycle. Establish a constraint set that includes linear and nonlinear constraints, the constraint set reflecting the physical operation limitations of the distributed photovoltaic energy storage area and the grid requirements; Robust optimization constraints are applied to address the uncertainty in photovoltaic power output forecasting; The sequential quadratic programming algorithm is used to solve the optimization model that includes the aforementioned nonlinear constraints and robust optimization constraints.

[0015] Preferably, the linear constraint includes: Energy storage charging and discharging power limitations: , in, For energy storage charging power, For energy storage discharge power, These are the maximum power limits for charging and discharging. Safe range of state of charge for energy storage: The battery state of charge is maintained between 5% and 95% of its maximum state of charge; Power grid interaction limits: The active power range for power grid interaction is [-500kW, 500kW], and the reactive power range is [-500kVar, 500kVar]. Uncertainty in photovoltaic processing: The uncertainty coefficient of photovoltaic output in each time period ranges from [0,1], and the sum of the uncertainty coefficients of all time periods is less than a set threshold; The nonlinear constraints are as follows: Power balance constraint: In the worst case, the sum of photovoltaic power generation and energy storage discharge power equals the active power consumed by the load, the energy storage charging power, and the active power transmitted to the grid; the sum of reactive power generated by photovoltaic and energy storage equals the reactive power consumed by the load and the reactive power transmitted to the grid. Equipment capacity constraints: The sum of the squares of the difference between the reactive power and the charge / discharge power of the energy storage system is less than or equal to the square of the maximum apparent power of the energy storage system; the sum of the squares of the active power and reactive power of the photovoltaic system is less than or equal to the square of the rated capacity of the photovoltaic inverter. Power factor constraint for power grid: The ratio of the absolute value of reactive power to the apparent power of the power grid is less than or equal to the value determined by the minimum power factor, as shown in the following formula: , in, This is the minimum power factor required by the power grid. Reactive power exchanged between the power grid and the grid. Active power exchanged between the power grid; State of charge (SCC) dynamic update constraint: The SCC at the current time point equals the SCC at the previous time point plus the SCC gained from charging, minus the SCC lost from discharging. Final value constraint of charged state: , in, This indicates the state of charge value at the end of the time. This represents the initial state of charge.

[0016] Preferably, the robust optimization constraints for the uncertainty of the photovoltaic power output prediction include: Obtain the short-term and long-term forecast ranges of photovoltaic power output at each forecast time step; For each time step, an uncertain decision variable with a value range of [0,1] is introduced; Constructing a set of budget uncertainties: , in, Represents decision variables under uncertainty. Indicates budget parameters; All constraints in the constraint set are robustly processed using a budget uncertainty set.

[0017] Preferably, the sequential quadratic programming algorithm is solved using the fmincon function in MATLAB, and the output variables include: the optimal charging and discharging power of the energy storage system, the reactive power of the energy storage system, and the optimal active / reactive power interacting with the power grid.

[0018] A distributed photovoltaic energy storage area active / reactive power support capability optimization device includes a processor and a memory storing program instructions. The processor is configured to execute the distributed photovoltaic energy storage area active / reactive power support capability optimization method when running the program instructions.

[0019] The advantages of this invention are as follows: The proposed EMD-KPCA-LSTM combined prediction model can fully explore the multi-scale, nonlinear correlation between meteorological data and photovoltaic power output, effectively remove data redundancy and noise, and significantly improve the accuracy and robustness of short-term photovoltaic power prediction, especially in dealing with complex weather fluctuations. The photovoltaic power prediction method of this invention can effectively quantify the uncertainty of prediction, provide confidence intervals for predicted values, and provide reliable data support for subsequent risk-based optimized scheduling.

[0020] The nonlinear programming model based on robust optimization constructed in this invention can effectively integrate the uncertainty information of photovoltaic power output into the optimization problem, making the optimization strategy more adaptable and reliable under photovoltaic fluctuations. The SQP algorithm used in this invention can effectively solve optimization models containing complex nonlinear constraints, ensuring the accuracy and practicality of the optimization results, and providing a feasible solution for the actual operation and control of distributed photovoltaic and energy storage areas.

[0021] By combining the aforementioned high-precision uncertainty prediction with robust nonlinear optimization, this invention can significantly improve the active / reactive power support capability of distributed photovoltaic-storage distribution areas to the power grid, and enhance the stability, flexibility, and renewable energy absorption capacity of the power grid. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0023] Figure 1 This is a schematic diagram of the photovoltaic power output prediction method of the present invention.

[0024] Figure 2 This is a schematic diagram of the process for optimizing the active / reactive power support capacity of the distributed photovoltaic energy storage area according to the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] This invention provides a photovoltaic power output prediction method; under the premise of the photovoltaic power output prediction value obtained by the photovoltaic prediction model and the quantified uncertainty information as input, it provides an optimization of the active / reactive power support capacity of a distributed photovoltaic and energy storage area.

[0027] Example 1 like Figure 1 As shown, a photovoltaic power output prediction method includes the following steps: S1: Clean historical photovoltaic power data and various environmental meteorological parameter data, process outliers and missing values, and form a training dataset; S2: Perform EMD decomposition on the environmental meteorological parameter data to generate a feature sequence composed of intrinsic mode functions and residual terms; S3: Use a multinomial kernel function to perform KPCA dimensionality reduction on the feature sequence while retaining the principal components; S4: Input the dimensionality-reduced data into the Long Short-Term Memory network for training, build a time series prediction model, and output the photovoltaic power prediction value; S5: Quantify the uncertainty of the prediction results, calculate the Z-score and standard deviation based on the predicted values, and generate the confidence interval of the predicted values.

[0028] As a refinement of the above embodiments, the collected historical photovoltaic power data and various environmental meteorological parameter data, such as air temperature, azimuth angle, cloud opacity, dew point temperature, solar diffuse radiation index (DHI), solar direct radiation index (DNI), total solar horizontal radiation (GHI), fixed tilt radiation (GTI), tracking tilt radiation (GTI), atmospheric precipitable water, relative humidity, snow depth, surface air pressure, wind direction at 10m altitude, wind speed at 10m altitude, zenith angle, etc., are cleaned to process outliers and missing values, forming a dataset suitable for model training.

[0029] As a refinement of the above embodiments, the empirical mode decomposition of the data includes: (1) Local extreme point location: It is necessary to determine the input signal. All local maxima and local minima are key elements for constructing the signal envelope.

[0030] (2) Envelope Generation: Using interpolation techniques, all local maxima and local minima are connected. Spline interpolation methods (e.g., cubic spline interpolation) are used to ensure the smoothness of the envelope, thereby obtaining the upper envelope of the signal. and lower envelope .

[0031] (3) Calculate the mean line Averaging the upper and lower envelopes yields the signal mean, which reflects the low-frequency trend of the signal. .

[0032] (4) Preliminary extraction of detail components: from the original signal Subtract the mean envelope from the middle This yields a candidate detail component. It is important to note that this detail is significant. It may not necessarily meet the IMF's definition, and therefore may require further screening.

[0033] Finally, through EMD decomposition, the 16 meteorological data were decomposed into a total of 158-dimensional feature sequences containing 142-dimensional IMF components and 16-dimensional residual components, and these were used as a new set of feature sequences.

[0034] As a refinement of the above embodiments, the step of performing KPCA dimensionality reduction on the feature sequence using a polynomial kernel function includes: This embodiment employs kernel principal component analysis (KPCA) for dimensionality reduction. A polynomial kernel function is selected, and its order is set to d=18. The data is mapped to a high-dimensional space of an 18th-degree polynomial, and then PCA dimensionality reduction is performed in this space. To ensure that the dimensionality-reduced data fully represents the original data, components with a cumulative contribution rate of 90% are selected as principal components. The selected principal components have strong representativeness of the original feature sequences and possess stronger comprehensive information capabilities. After KPCA dimensionality reduction, the data dimension is reduced to 26 dimensions.

[0035] As a refinement of the above embodiments, the method of using an LSTM model for photovoltaic power prediction includes: In this embodiment, a sample set containing 5200 time-section data points is transformed into a data format suitable for LSTM network training, and divided into training and test sets in a 7:3 ratio. During the training of the LSTM network, this embodiment uses the Adaptive Moment Estimation (Adam) optimization algorithm to update the network weights, and adaptively adjusts the learning rate based on the gradient information of each parameter. This embodiment finely tunes the model parameters, and the final parameter settings are as follows: (1) Input layer time steps: 5; (2) Input layer dimension: 9; (3) Number of hidden layers: 1; (4) Number of hidden layer units: 100; (5) Output layer dimension: 1; (6) Training batch size: 25; (7) Number of training epochs: 100; (8) Number of training set samples: 3640; (9) Number of test set samples: 1560.

[0036] As a refinement of the above embodiments, the step of quantifying the uncertainty of the prediction results includes: This invention further calculates the 95% confidence interval of the photovoltaic output power prediction values ​​for two days of the EMD-KPCA-LSTM prediction model test set by calculating the Z-scor and the standard deviation of the residual, and judges the accuracy by the interval coverage and average width.

[0037] It should be noted that the proposed EMD-KPCA-LSTM combined prediction model can fully explore the multi-scale, nonlinear correlation between meteorological data and photovoltaic power output, effectively remove data redundancy and noise, and significantly improve the accuracy and robustness of short-term photovoltaic power prediction, especially in dealing with complex weather fluctuations. The photovoltaic power prediction method of this invention can effectively quantify the uncertainty of prediction, provide confidence intervals for predicted values, and provide reliable data support for subsequent risk-based optimized scheduling.

[0038] Example 2 like Figure 2 As shown, a method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area includes the following steps: S1: Obtain the photovoltaic power prediction value based on the photovoltaic output prediction method, and quantify the uncertainty of the prediction result; S2: The objective function is set to maximize the sum of the absolute values ​​of active power and reactive power of the interaction between the distributed photovoltaic energy storage area and the main power grid within the scheduling cycle. S3: Establish a constraint set that includes linear and nonlinear constraints, the constraint set reflecting the physical operation limitations of the distributed photovoltaic energy storage area and the grid requirements; S4: Robust optimization constraints are applied to address the uncertainty in photovoltaic power output prediction; S5: Use the sequential quadratic programming algorithm to solve the optimization model that includes the nonlinear constraints and robust optimization constraints.

[0039] As a refinement of the above embodiments, the objective function is set to maximize the sum of the absolute values ​​of active power and reactive power interacting between the distributed photovoltaic energy storage area and the main power grid during the scheduling period, including: , in, This represents the active power delivered to the power grid at time t (positive for output, negative for input). This represents the reactive power delivered to the grid at time t (positive represents inductive reactive power, negative represents capacitive reactive power). Indicates the end time.

[0040] As a refinement of the above embodiments, the linear constraints and nonlinear constraints include: 1. Linear constraints (1) Energy storage charging and discharging power limit: The charging power and discharging power of the energy storage battery must both be greater than or equal to 0 and less than or equal to the maximum power. Furthermore, the difference between charging power and discharging power, as well as the difference between discharging power and charging power, must both be less than or equal to the maximum power. .

[0041] , in, For energy storage charging power, For energy storage discharge power, This refers to the maximum power limit for charging and discharging.

[0042] (2) Safe range of energy storage state of charge: The battery's state of charge (SOC) must be maintained at the maximum state of charge. Between 5% and 95%.

[0043] , in, The energy storage state of charge is at time t.

[0044] (3) Power exchange limit with the grid: Active power exchanged with the grid and reactive power Both have upper and lower limits, with ranges of ±500kW and 500kVar respectively.

[0045] , (4) Photovoltaic output uncertainty: The photovoltaic output uncertainty coefficient for each time period is between 0 and 1, and the sum of the uncertainty coefficients for all time periods does not exceed a given total scale limit.

[0046] , in, The normalization coefficient for the uncertainty of photovoltaic power output. This represents the budget parameter, which restricts the uncertainty parameter (here) across all time steps. This parameter indicates how many units the uncertainty can "move" to its worst-case value in total. It reflects that while individual uncertainty parameters may fluctuate within a given range, it is rare for all uncertainty parameters to simultaneously reach their worst-case extreme values. This is achieved by setting... This allows for a balance between worst-case performance guarantees and more general (but possibly not absolute worst-case) performance. In this embodiment, The value is 20.

[0047] 2. Nonlinear constraints (1) Power balance constraint: Photovoltaic power generation in the worst case With energy storage discharge power The sum equals the active power consumed by the load. Energy storage charging power and the active power transmitted to the power grid Reactive power generated by photovoltaics and reactive power generated by energy storage The sum equals the reactive power consumed by the load. and reactive power transmitted to the power grid .

[0048] , , , , in: This represents the lower bound of the active power within the confidence interval for photovoltaic power prediction. Let represent an uncertain decision variable with values ​​ranging from [0,1], used for robust optimization. This represents the upper bound of the active power within the confidence interval for photovoltaic power prediction. This represents the minimum power factor of the photovoltaic inverter, with a value of 0.9.

[0049] Worst-case scenario for photovoltaic power output ( () refers to a point in time. tThis refers to the minimum actual photovoltaic power generation that can be achieved, taking into account the uncertainty of photovoltaic power generation forecasts. This minimum value is obtained by correcting the photovoltaic power forecast value, taking into account forecast errors and uncertainty factors. It refers to the output configuration that is most unfavorable to the current optimization objective (maximizing power support) or constraints (such as power balance and capacity limitations) among all possible photovoltaic output combinations within the forecast range.

[0050] (2) Equipment capacity constraints: The sum of the squares of the difference between the reactive power and the charge / discharge power of the energy storage system shall be less than or equal to the square of the maximum apparent power of the energy storage system; the sum of the squares of the photovoltaic active power and the photovoltaic reactive power in the worst case shall be less than or equal to the square of the rated capacity of the photovoltaic inverter.

[0051] , in, For energy storage apparent power, This refers to the capacity of the photovoltaic inverter.

[0052] (3) Power factor constraint of the power grid: The ratio of the absolute value of the reactive power of the power grid to the apparent power of the power grid is less than or equal to a value determined by the minimum power factor required by the power grid.

[0053] , in, The minimum power factor required by the power grid is 0.9.

[0054] (4) Dynamic update of state of charge: State of charge at the first time point equal to the initial state of charge Adding the increase in state of charge due to charging and subtracting the decrease in state of charge due to discharging, starting from the second time point, the state of charge at each time point is equal to the state of charge at the previous time point. In addition to the increase in state of charge due to charging, the decrease in state of charge due to discharging is subtracted.

[0055] , in, The charge / discharge efficiency is 95%.

[0056] (5) Final value constraint of state of charge: , in, This represents the SOC value at the end of the time period. This represents the initial SOC value, ensuring that the stored energy returns to its initial state at the end of the day.

[0057] As a refinement of the above embodiments, the robust optimization constraints for the uncertainty of photovoltaic power output prediction include: This method first obtains the short-term and long-term forecast intervals of photovoltaic power output at each forecast time step, which are bounded by a lower bound. and an upper realm Definition. To quantify uncertainty and handle it in the optimization model, an uncertainty decision variable is introduced for each time step t. Its value range is constrained to [0,1]. Based on this, a worst-case model is constructed to represent the photovoltaic output considering uncertainties.

[0058] To avoid overly conservative optimization results and to reflect the fact that uncertainties in reality do not reach their worst-case values ​​simultaneously at all time steps, a budget uncertainty set is adopted. Constraining uncertain variables. Specifically, this involves introducing a budget parameter and imposing constraints. This constraint limits the sum of uncertainties at all time steps, thereby controlling the overall level of uncertainty.

[0059] Subsequently, when constructing the optimal scheduling model for distributed photovoltaic and energy storage areas, the key constraints related to photovoltaic output were robustly processed. These constraints were reformulated to ensure that they must be satisfied for any combination of uncertainties that satisfy the aforementioned set of budget uncertainties. In this way, the optimization problem was transformed into finding a scheduling strategy that is feasible under all possible "worst-case" photovoltaic output scenarios.

[0060] As a refinement of the above embodiments, the step of solving the model using a sequential quadratic programming algorithm includes: The solution process of the Sequential Quadratic Programming (SQP) algorithm mainly involves two steps: First, the original nonlinear optimization problem is transformed into a simplified quadratic optimization problem through Taylor expansion; second, this quadratic optimization problem is solved iteratively. The iterative solution steps are as follows: (1) Given an initial point Convergence accuracy ,make , place ; (2) The original problem at the iteration point This can be simplified into a quadratic programming problem. (3) Solve the above quadratic programming problem and let ; (4) In direction The objective function of the original problem is constrained and searched in one dimension to obtain the next iteration point. ; (5) Termination judgment: If If the termination criterion of a given precision is met, then... As the optimal solution. If the optimal cost is the objective function, terminate the calculation; otherwise, go to (6). (6) Corrected according to DFP or BFGS method ,make , turn (2).

[0061] This represents an approximation of the initial Hessian matrix. In the SQP algorithm, the Hessian matrix (or its approximation) is used to describe the curvature information of the objective function, thereby guiding the search direction. This is the initial estimate of the Hessian matrix at the start of the iterative process. Typically, without better prior information, It will be set as the identity matrix . This represents the identity matrix, which is a diagonal matrix where all diagonal elements are 1s and all other elements are 0s. In the SQP algorithm, this is used to... Initializing to an identity matrix implies assuming that the objective function has a uniform curvature during the initial iteration. This indicates the number of iterations. The SQP algorithm is an iterative process that gradually approaches the optimal solution through multiple iterations. Used to record the current iteration count, after each iteration. The value will increase by 1. The optimal solution to the quadratic programming subproblem is represented by the SQP algorithm, which approximates the original nonlinear programming problem as a quadratic programming (SQP) subproblem in each iteration. It is the search direction obtained after solving the QP subproblem, indicating the direction from the current iteration point to the next iteration point.

[0062] Finally, the quadratic sequence programming algorithm was used to optimize the model using the fmincon function in MATLAB, and the optimal values ​​of the energy storage system's charging and discharging power, energy storage reactive power, and optimal active / reactive power for interaction with the grid were obtained.

[0063] It should be noted that the robust optimization-based nonlinear programming model constructed in this embodiment can effectively integrate the uncertainty information of photovoltaic power output into the optimization problem, making the optimization strategy more adaptable and reliable under photovoltaic fluctuations. The SQP algorithm used in this invention can effectively solve optimization models containing complex nonlinear constraints, ensuring the accuracy and practicality of the optimization results, and providing a feasible solution for the actual operation and control of distributed photovoltaic energy storage areas.

[0064] Example 3 This disclosure also provides a detection device for standardizing the wearing of judges' robes in court, including a processor and a memory. Optionally, the device may further include a communication interface and a bus. The processor, communication interface, and memory can communicate with each other via the bus. The communication interface can be used for information transmission. The processor can call logical instructions in the memory to execute the detection method for standardizing the wearing of judges' robes in court, as described in the above embodiment.

[0065] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0066] Memory, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as the program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor executes the program instructions / modules stored in the memory to perform functional applications and data processing, thereby implementing the detection method for ensuring the proper wearing of judges' robes in court as described in the above embodiments.

[0067] The memory may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory may include high-speed random access memory and may also include non-volatile memory.

[0068] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code. It can also be a transient storage medium.

[0069] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting photovoltaic power output, characterized in that, Includes the following steps: Historical photovoltaic power data and various environmental meteorological parameter data are cleaned, outliers and missing values ​​are processed, and a training dataset is formed. The environmental meteorological parameter data are subjected to EMD decomposition to generate a feature sequence composed of intrinsic mode functions and residual terms; The feature sequence is reduced in dimensionality using a multinomial kernel function while retaining principal components; The dimensionality-reduced data is input into a long short-term memory network for training, a time series prediction model is constructed, and the photovoltaic power prediction value is output. The uncertainty of the prediction results is quantified, and the confidence interval of the prediction value is generated based on the calculated Z-score and standard deviation of the prediction value.

2. The photovoltaic power output prediction method according to claim 1, characterized in that, The environmental meteorological parameters include: air temperature, azimuth, cloud opacity, dew point temperature, solar diffuse radiation index, solar direct radiation index, total horizontal solar radiation, fixed tilt radiation, tracking tilt radiation, atmospheric precipitable water, relative humidity, snow depth, surface air pressure, wind direction at 10m altitude, wind speed at 10m altitude, and zenith angle.

3. The photovoltaic power output prediction method according to claim 1, characterized in that, The specific process of EMD decomposition includes: Locate the local maxima and minima of the input signal; The upper and lower envelopes are generated using spline interpolation. The mean line of the signal is obtained by averaging the upper and lower envelopes. : ， in, Indicates the upper envelope. Indicates the lower envelope; Subtract from the original signal Extract the detail components and iterate to obtain the IMF components and residual terms; The output is a 158-dimensional feature sequence consisting of 142-dimensional IMF components and 16-dimensional residual terms.

4. The photovoltaic power output prediction method according to claim 1, characterized in that, The KPCA uses a polynomial kernel function of order d=18, resulting in a data dimension of 26 after dimensionality reduction.

5. The photovoltaic power output prediction method according to claim 1, characterized in that, The LSTM model network structure parameters are as follows: input layer time step 5, input dimension 9, single hidden layer, number of hidden layer units 100, output dimension 1; The Adam optimizer is used to update the weights, with a batch size of 25 and 100 training epochs.

6. A method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area, characterized in that, Includes the following steps: Photovoltaic power prediction values ​​are obtained based on the photovoltaic output prediction method described in any one of claims 1-5, and the uncertainty of the prediction results is quantified. The objective function is set to maximize the sum of the absolute values ​​of active power and reactive power of the interaction between the distributed photovoltaic energy storage area and the main power grid within the scheduling cycle. Establish a constraint set that includes linear and nonlinear constraints, the constraint set reflecting the physical operation limitations of the distributed photovoltaic energy storage area and the grid requirements; Robust optimization constraints are applied to address the uncertainty in photovoltaic power output forecasting; The sequential quadratic programming algorithm is used to solve the optimization model that includes the aforementioned nonlinear constraints and robust optimization constraints.

7. The method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area according to claim 6, characterized in that, The linear constraints include: Energy storage charging and discharging power limitations: ， in, For energy storage charging power, For energy storage discharge power, These are the maximum power limits for charging and discharging. Safe range of state of charge for energy storage: The battery state of charge is maintained between 5% and 95% of its maximum state of charge; Power limits for grid interaction: The active power range for grid interaction is [-500kW, 500kW], and the reactive power range is [-500kVar, 500kVar]. Uncertainty in photovoltaic processing: The uncertainty coefficient of photovoltaic output in each time period ranges from [0,1], and the sum of the uncertainty coefficients of all time periods is less than a set threshold; The nonlinear constraints are as follows: Power balance constraint: In the worst case, the sum of photovoltaic power generation and energy storage discharge power equals the active power consumed by the load, the energy storage charging power, and the active power transmitted to the grid; the sum of reactive power generated by photovoltaic and energy storage equals the reactive power consumed by the load and the reactive power transmitted to the grid. Equipment capacity constraints: The sum of the squares of the difference between the reactive power and the charge / discharge power of the energy storage system is less than or equal to the square of the maximum apparent power of the energy storage system; the sum of the squares of the active power and reactive power of the photovoltaic system is less than or equal to the square of the rated capacity of the photovoltaic inverter. Power factor constraint for power grid: The ratio of the absolute value of reactive power to the apparent power of the power grid is less than or equal to the value determined by the minimum power factor, as shown in the following formula: ， in, The minimum power factor required by the power grid. Reactive power exchanged between the power grid and the grid. Active power exchanged between the power grid; State of charge (SOC) dynamic update constraint: The SOC at the current time point equals the SOC at the previous time point plus the SOC gained from charging, minus the SOC lost from discharging. Final value constraint of charged state: , in, This indicates the state of charge value at the end of the time. This represents the initial state of charge.

8. The method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area according to claim 6, characterized in that, The robust optimization constraints for the uncertainty of the photovoltaic power output prediction include: Obtain the short-term and long-term forecast ranges of photovoltaic power output at each forecast time step; For each time step, an uncertain decision variable with a value range of [0,1] is introduced; Constructing a set of budget uncertainties: ， in, Represents decision variables under uncertainty. Indicates budget parameters; All constraints in the constraint set are robustly processed using a budget uncertainty set.

9. The method for optimizing the active / reactive power support capacity of a distributed photovoltaic energy storage area according to claim 6, characterized in that, The sequential quadratic programming algorithm is solved using the fmincon function in MATLAB. The output variables include the optimal charging and discharging power of the energy storage system, the reactive power of the energy storage system, and the optimal active / reactive power interacting with the power grid.

10. A distributed photovoltaic energy storage area active / reactive power support capability optimization device, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the active / reactive power support capability optimization method for distributed optical energy storage areas as described in any one of claims 6-9 when running the program instructions.

Images