Photovoltaic capacity calculation method and system based on data driving and neural network

Abstract

The invention discloses a photovoltaic capacity calculation method and system based on data driving and a neural network. The method comprises the following steps: collecting and preprocessing historical operation data of a transformer area; carrying out feature extraction to construct an input feature vector, wherein the feature vector comprises a statistical feature extracted from the power data and a coding feature reflecting a time period characteristic; training a neural network by using the feature vectors and the corresponding voltage data, and establishing a data-driven transformer area voltage evaluation model; and setting voltage and load rate constraint conditions of transformer area operation, carrying out iterative simulation on a photovoltaic grid-connected scene based on the model and carrying out safety verification, and carrying out backstepping calculation to obtain the maximum photovoltaic capacity which can be opened and accessed in the transformer area. According to the method, the problem of photovoltaic capacity evaluation under the condition that the parameters of the power distribution network are missing or inaccurate is solved, the evaluation precision is high, the practicability is high, batch parallel calculation can be realized depending on an enterprise platform, and core decision support is provided for large-scale distributed photovoltaic safe grid connection and efficient consumption.

Description

Technical Field

[0001] This invention belongs to the field of new energy consumption and smart distribution network technology in power systems, and particularly relates to a photovoltaic capacity calculation method and system based on data-driven and neural networks. Background Technology

[0002] With the rapid development of county-wide and rooftop photovoltaic (PV) systems, the grid-connected capacity of distributed PV is constantly increasing. Given the current requirement to connect all distributed PV systems to existing grids, large-scale grid construction and upgrades are inevitable to address the safety issues arising from unrestricted distributed PV access. Therefore, scientifically and rationally assessing the available capacity of distributed PV in a given area and controlling the grid-connected capacity within that capacity is crucial for promoting the coordinated development of distributed PV and the distribution network.

[0003] A series of exploratory studies have been carried out both domestically and internationally on the algorithm for the open capacity of distributed photovoltaic (PV) in transformer substations. The main idea is to substitute the grid-connected capacity of each node into the existing power flow calculation model for calculation, and obtain the open capacity of distributed PV in transformer substations with the voltage not exceeding the limit. However, power flow calculation is based on the condition that the grid parameters are accurate and complete. However, in practical applications, the grid parameters of low-voltage distribution networks are basically missing, and the model building process is complicated, which limits the applicability in scenarios where low-voltage distribution network parameters are missing.

[0004] Therefore, this invention proposes a distributed photovoltaic capacity calculation method based on data-driven and neural network substation models to address the problem of power flow calculation being impossible due to missing grid parameters. Artificial neural network algorithms have become powerful tools for electricity demand forecasting because they can model complex nonlinear relationships in energy data. These models are particularly effective in dynamic environments such as shopping malls and residential buildings, where traditional statistical methods based on linear regression often struggle to capture complex patterns in energy consumption. Machine learning and deep learning models, by learning from historical data and discovering patterns influencing future demand, achieve higher accuracy in predicting energy consumption. Summary of the Invention

[0005] To address these issues, this invention provides a photovoltaic capacity calculation method based on data-driven approaches and neural networks, which solves the aforementioned problems.

[0006] In a first aspect, the present invention provides a photovoltaic capacity calculation method based on data-driven and neural networks, comprising: S1. Collect historical operating data of the transformer area and preprocess the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer area outlet voltage data and load rate data; S2. Extract features from the standardized data to construct a feature vector containing power features and timing features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. S3. Set safety and stability constraints for the operation of the distribution area. Based on the voltage evaluation model of the distribution area, iteratively simulate the photovoltaic grid connection scenario and perform stability verification to deduce the maximum photovoltaic capacity that the distribution area can open.

[0007] Secondly, the present invention provides a photovoltaic capacity calculation system based on data-driven and neural networks, comprising: The data acquisition and preprocessing module is configured to acquire historical operating data of the transformer area and preprocess the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer area outlet voltage data, and load rate data; The feature extraction and model training module is configured to extract features from the standardized data, construct a feature vector containing power features and time-series features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. The capacity calculation and reverse calculation module is configured to set safety and stability constraints for the operation of the transformer area, and based on the transformer area voltage evaluation model, it iteratively simulates the photovoltaic grid connection scenario and performs stability verification to reverse calculate the maximum photovoltaic capacity that the transformer area can open. The big data platform interface module is configured to rely on the big data cluster of the enterprise middle platform to modularize and standardize the data processing, model calling and capacity calculation process; using a distributed computing framework, steps S1 to S3 are executed in parallel for multiple transformer areas to achieve batch photovoltaic capacity assessment.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the photovoltaic capacity calculation method based on data-driven and neural networks according to any embodiment of the present invention.

[0009] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the steps of the photovoltaic capacity calculation method based on data-driven and neural networks according to any embodiment of the present invention.

[0010] The photovoltaic capacity calculation method and system based on data-driven and neural networks in this application have the following specific advantages: 1) Breakthrough in overcoming the bottleneck of grid parameter dependence: By adopting a pure data-driven approach, the complex mapping relationship between power and voltage is learned from historical operating data using neural networks, completely avoiding dependence on precise grid physical parameters and solving the industry problem of being unable to conduct capacity assessment under the condition of missing low-voltage distribution network parameters.

[0011] 2) Improved assessment accuracy and efficiency: The neural network model can capture nonlinear and temporal characteristics, resulting in high assessment accuracy. Combined with the distributed parallel computing capabilities of the enterprise middleware platform, the originally time-consuming and laborious single-point calculations can be transformed into efficient batch assessments, greatly improving the efficiency of large-scale photovoltaic planning and management of distribution networks. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a photovoltaic capacity calculation method based on data-driven and neural networks, provided as an embodiment of the present invention; Figure 2 A structural block diagram of a photovoltaic capacity calculation system based on data-driven and neural networks is provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0015] Please see Figure 1 The diagram shows a flowchart of a photovoltaic capacity calculation method based on data-driven and neural networks according to this application.

[0016] In one embodiment, step S1 involves collecting historical operating data of the transformer substation and preprocessing the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer substation outlet voltage data, and load rate data. Specifically, historical operating data of the transformer substation is collected by a high-speed power line carrier communication HPLC smart meter or concentrator installed on the low-voltage side of the transformer substation. The sampling interval is 15 minutes, and the data time span is not less than 1 year.

[0017] Specifically, the historical operational data is preprocessed, including: Use linear interpolation to fill in missing data; For data segments with more than 4 consecutive missing sampling points, the average curve of the same type of date is used for replacement; Outliers are identified based on the 3σ criterion and replaced with the moving average of adjacent normal data. The net load, voltage, and load factor data are normalized to the [0,1] interval.

[0018] In this step, historical operating data for one year was collected using an HPLC smart meter installed on the low-voltage side of the transformer in this distribution area. The collected data includes: Net load active power time series: sampling interval is 15 minutes, 96 points per day. Positive values ​​indicate power supply from the grid to users, and negative values ​​indicate power feedback from users to the grid.

[0019] Distribution area outlet voltage data: outlet voltage values ​​collected synchronously.

[0020] Line load rate and distribution transformer load rate data.

[0021] Preprocess the collected raw data: Missing value handling: For missing data points, either a single point or 2-3 consecutive points, linear interpolation is used for imputation. For data segments with more than 4 consecutive missing points (e.g., due to communication interruption), average load curves from the same date type (e.g., all Tuesdays) are used for replacement.

[0022] Outlier handling: Outlier data is identified based on the 3σ criterion. The mean (μ) and standard deviation (σ) of the power data at each time point are calculated. Data points that exceed the range of (μ-3σ, μ+3σ) are identified as outliers and replaced with the moving average of adjacent normal data.

[0023] Data normalization: Normalize all net load, voltage and load factor data to the [0, 1] range to eliminate dimensional differences.

[0024] The specific method for obtaining the average curve is as follows: The week is divided into three groups. The first group, Monday to Thursday, is called the typical weekday group. During this period, residents have regular schedules, working or studying during the day, resulting in a typical "double-peak" electricity consumption pattern. The second group, Friday to Saturday, is called the social / late-night day. During this period, residents have more social activities and often return home late, causing the evening peak electricity consumption to be delayed, weakened, or flattened. The third group is Sunday, called the stay-at-home day. During this period, residents are mostly at home, resulting in higher daytime electricity demand than on weekdays, and earlier bedtimes. The load curve shape is between that of weekdays and Friday / Saturday, but it has its own unique characteristics.

[0025] Then, average the data. Collect 24-hour net load time series data from the transformer substation master table for several consecutive days. P net =[ p 1 , p 2 ,..., p T ],in T For each daily data point, the raw data is cleaned, including handling missing values ​​and removing obvious outliers. Based on the date corresponding to each net load curve, it is categorized into one of the three groups mentioned above: GroupA = {all data curves for Monday, Tuesday, Wednesday, and Thursday}; GroupB = {Data curves for all Fridays and Saturdays}; GroupC = {all Sunday data curves}, which is the arithmetic mean of all net load curves within each group, calculated at different time points. This generates an "average load curve" that represents the typical pattern of this group.

[0026] Step S2: Extract features from the standardized data to construct a feature vector containing power features and time-series features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. Specifically, the neural network model is an artificial neural network comprising one input layer, two hidden layers, and one output layer; The number of neurons in the input layer is the same as the dimension of the feature vector; Each of the hidden layers contains 32 neurons, and the activation function is the ReLU function; The output layer contains one neuron, which is used to output the predicted voltage value of the station area; The neural network model employs the Adam optimizer, with the goal of minimizing the root mean square error (RMSE) during training, and uses early stopping to prevent overfitting.

[0027] In this step, a neural network model is trained based on historical data to achieve a precise mapping from power characteristics to voltage. The feature vector includes power statistical features (such as daily power mean and standard deviation) and time-series encoded features (such as season, weekday date type, and electricity price period).

[0028] In this embodiment, the feature vector includes 12 power features and 3 timing features; The power characteristics include: daily power average, standard deviation, peak-to-valley difference, power percentage during the daytime (9:00-17:00), and power percentage during the nighttime (23:00-5:00). The temporal features include: seasonal features, intra-week grouping features, and time-period features; among which... The seasonal characteristics are coded as 1, 2, and 3 for summer, winter, and spring / autumn, respectively. The weekday grouping features are coded as 1, 2, and 3 for Monday to Thursday, Friday to Saturday, and Sunday, respectively. The time period characteristics are coded as 1, 2, and 3 for peak periods, flat periods, and low periods, respectively.

[0029] First, perform feature engineering: S21. Extract 12 power characteristics. Calculate the daily power mean, standard deviation, peak-to-valley difference, daytime (9:00-17:00) power ratio, and nighttime (23:00-5:00) power ratio from the daily net load curve.

[0030] S22. Add 3 timing features.

[0031] Seasonal characteristics: Summer (June-August) is coded as 1, winter (December-February) is coded as 2, and spring and autumn (March-May, September-November) are coded as 3.

[0032] Weekday grouping characteristics: Monday to Thursday (typical workdays) are coded as 1, Friday to Saturday (social / late-night days) are coded as 2, and Sunday (stay-at-home days) are coded as 3.

[0033] Time period characteristics: According to the power grid pricing policy, peak hours are coded as 1, average hours as 2, and off-peak hours as 3.

[0034] S23. Combine the above 15 features into a 15-dimensional feature vector, and use it as the input to the model.

[0035] Then train the model: S24. An artificial neural network (ANN) architecture is adopted, specifically: an input layer (15 neurons), two hidden layers (32 neurons each), and an output layer (1 neuron).

[0036] S25. The hidden layer uses the ReLU function as the activation function, and the output layer uses the linear activation function.

[0037] S26. Use 70% of the data from the whole year as the training set, 20% as the validation set, and 10% as the test set.

[0038] S27. Using historical power feature vectors as input and corresponding historical voltage data as labels, the Adam optimizer is used for training, with the goal of minimizing the root mean square error (RMSE).

[0039] S28. Early stopping is employed to prevent overfitting. Training is automatically terminated when the validation set loss no longer decreases for five consecutive epochs. The final training yields a voltage evaluation model with high accuracy and strong generalization ability.

[0040] It should be noted that an artificial neural network is a machine learning algorithm designed to learn a nonlinear function from a set of input values ​​to output values, as shown below: ; In the formula It is a nonlinear function parameterized by ANN. Let M be the input feature set, and Y be the output value set. Each set of input features m contains I samples, and each sample is represented as... These samples constitute the training dataset, which can be described as follows: ; Where D is the training dataset.

[0041] The basic structure of an artificial neural network consists of three main components: the input layer, the hidden layers, and the output layer. The input layer contains a set of neurons, each representing an input feature value. Each hidden layer contains a set of neurons that transform the values ​​from the previous layer using the following function.

[0042] ; In the formula and These are the output and input of neuron n in each hidden layer, respectively, and σ is a non-linear activation function. These are the weights of neurons n in each hidden layer. This is the bias term. The ReLU function is chosen as the activation function because it preserves linear properties and simplifies the gradient optimization process.

[0043] The output layer has a single neuron that directly outputs the estimated PV capacity value without applying an activation function. The output is calculated via forward propagation, and the weights are updated via backpropagation to minimize the loss function. Backpropagation utilizes the gradient descent principle to iteratively adjust the parameters, minimizing the prediction error.

[0044] The advantages of ANNs lie in their ability to learn complex nonlinear relationships, update the model in real time, and be insensitive to local minima. However, their disadvantages include the randomness of weight initialization, sensitivity to scaling of input features, and the need for manual tuning of hyperparameters.

[0045] Therefore, the specific method for using the Adam optimizer to optimize hyperparameters, including learning rate and regularization parameters, is as follows: The Adam optimizer is used as the core optimization algorithm for training artificial neural networks. Because it combines the advantages of momentum method and RMSProp, it adaptively adjusts the learning rate of each parameter by calculating the exponential moving average of the first moment (mean) and second moment (variance) of the gradient, effectively handling non-stationary gradients and accelerating the convergence process.

[0046] Hyperparameter optimization includes the initial learning rate, L2 regularization parameter, and maximum number of iterations. The initial learning rate and L2 regularization parameter are optimized using a grid search technique. Multiple iterations are performed on the learning rate within a set interval (in this embodiment, the interval is set to [0.001, 0.01, 0.1]), and the optimal initial learning rate is finally determined to be 0.01. Simultaneously, the L2 regularization parameter is tested within a set range (in this embodiment, the range is set to

[10] ). -5 5×10 -5 10 -4 The optimal value was determined to be 5 × 10. -5 This process continues until the same parameter combination appears twice consecutively. The optimization of the maximum number of iterations is achieved by real-time monitoring of the root mean square error (RMSE) of the training and validation sets. When the RMSE of the validation set no longer decreases for five consecutive iterations, an early stopping mechanism is triggered to prevent overfitting. In this embodiment, the maximum number of iterations is set to 210.

[0047] Step S3: Set safety and stability constraints for the operation of the distribution area. Based on the voltage evaluation model of the distribution area, iteratively simulate the photovoltaic grid connection scenario and perform stability verification to deduce the maximum photovoltaic capacity that the distribution area can open.

[0048] Specifically, the safety and stability constraints include: the voltage deviation at the transformer substation outlet does not exceed ±7% of the rated voltage, the load rate of the main line does not exceed 80%, and the load rate of the distribution transformer does not exceed 85%. The iterative simulation of photovoltaic grid-connected scenarios includes: S31: Set the initial value of photovoltaic capacity based on the number of users in the distribution area; S32: Input the initial value into the photovoltaic power prediction model to obtain the photovoltaic output curve, and superimpose it with the base load curve of the distribution area to generate a simulated net load curve after grid connection; S33: Input the simulated net load curve into the transformer substation voltage assessment model to obtain the voltage prediction value and calculate the line and transformer load rate; S34: Verify whether the predicted voltage value and the load rate meet the safety and stability constraints; If the conditions are met, increase the photovoltaic capacity and repeat steps S32-S34; If the requirements are not met, reduce the photovoltaic capacity and repeat steps S32-S34; Until the maximum photovoltaic capacity that satisfies all constraints is found.

[0049] In this step, the method further includes step S4: Based on the big data cluster of the enterprise middle platform, the data processing, model calling and capacity calculation process is modularized and standardized; using a distributed computing framework, steps S1 to S3 are executed in parallel for multiple transformer areas to achieve batch photovoltaic capacity assessment.

[0050] In this step, the first step is to set the constraint thresholds. Based on the "Distribution Network Operation Regulations" and the actual parameters of the distribution area, the stability constraint thresholds are set as follows: Voltage deviation at the transformer substation outlet: not exceeding ±7% of the rated voltage (0.4kV), meaning the operating voltage range must be maintained between 0.372kV and 0.428kV.

[0051] Main line load rate: ≤80%.

[0052] Distribution transformer load rate: ≤80% in summer, ≤85% in other seasons.

[0053] Then, perform reverse calculations to estimate the photovoltaic capacity: Initialization: Set the initial photovoltaic capacity trial value based on the number of users in the transformer area.

[0054] Simulated grid connection: The current test capacity value is substituted into a pre-trained photovoltaic power prediction model. The photovoltaic power prediction model is trained based on historical sunshine data and uses the LightGBM algorithm. The inputs are seasonal type, daily type, time period code, and meteorological data (the meteorological data is obtained from the enterprise's meteorological station, including irradiance, ambient temperature, and cloud cover, with the time resolution consistent with the power data). The output is a typical daily photovoltaic output curve. This photovoltaic output curve is superimposed on the original base load curve of the distribution area to generate a simulated "net load curve after photovoltaic grid connection".

[0055] Safety Assessment: The simulated net load curve is input into the voltage assessment model trained in step S102 to predict the voltage values ​​at each time point. Simultaneously, based on line and transformer parameters, the corresponding line load rate and distribution transformer load rate are calculated.

[0056] Iterative optimization: Verify whether the predicted voltage and load factor all meet the above constraint thresholds. If all are met, increase the photovoltaic capacity trial value (e.g., increase by 10kW) and jump back to step b to resimulate. If any indicator is not met, decrease the photovoltaic capacity trial value (e.g., decrease by 10kW) and jump back to step b to resimulate.

[0057] Through the above iterative approximation process, a maximum photovoltaic capacity value is finally found, ensuring that the operational indicators at all time points after grid connection meet safety and stability constraints. This capacity value is the available photovoltaic capacity for that distribution area.

[0058] Specifically, the method also includes step S4, which uses the big data cluster of the enterprise middle platform to modularize and standardize the data processing, model calling and capacity calculation process; and uses a distributed computing framework to execute steps S1 to S3 in parallel for multiple transformer areas to achieve batch photovoltaic capacity assessment.

[0059] In this step, batch processing is performed using the enterprise's middleware big data cluster.

[0060] First, the middle platform is adapted. The preprocessed data from each transformer area is standardized and stored in the middle platform's distributed database. The trained voltage assessment model and capacity back-calculation algorithm are encapsulated into standardized algorithm modules and deployed to the middle platform's model repository, supporting the configuration of constraint thresholds and the number of users for different transformer areas through parameters.

[0061] Then, batch parallel computing is performed. Utilizing the Spark distributed computing framework of the middle platform, the list of transformer substations to be computed is distributed to different computing nodes according to administrative regions. Each node performs data preprocessing, voltage assessment model invocation, and capacity back-calculation in parallel for the assigned transformer substations.

[0062] Final Results Output. After calculation, the platform automatically generates a photovoltaic capacity assessment report for each distribution area, including key indicators such as final capacity value, maximum voltage deviation, and peak load rate. All results are centrally displayed through the platform's BI visualization tools, supporting various formats such as maps, charts, and reports, for managers to query, export, and use for decision-making.

[0063] Using this method to evaluate the above example transformer area, the calculation is completed in just a few minutes. Compared with the traditional power flow calculation method that requires complete grid parameters, the efficiency is greatly improved, and the bottleneck of missing parameters is effectively avoided.

[0064] Please see Figure 2 The diagram shows a structural block diagram of a photovoltaic capacity calculation system based on data-driven and neural networks according to this application.

[0065] like Figure 2 As shown, the module includes a data acquisition and preprocessing module 200, a feature extraction and model training module 201, a capacity calculation and reverse calculation module 202, and a big data platform interface module 203.

[0066] The data acquisition and preprocessing module 200 is configured to acquire historical operating data of the transformer area and preprocess the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer area outlet voltage data and load rate data. The feature extraction and model training module 201 is configured to extract features from the standardized data, construct a feature vector containing power features and time-series features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. The capacity calculation and reverse calculation module 202 is configured to set safety and stability constraints for the operation of the transformer area, and based on the transformer area voltage evaluation model, iteratively simulates the photovoltaic grid connection scenario and performs stability verification to reverse calculate the maximum photovoltaic capacity that the transformer area can open. The big data platform interface module 203 is configured to rely on the big data cluster of the enterprise middle platform to modularize and standardize the data processing, model calling and capacity calculation process; using a distributed computing framework, the steps S1 to S3 are executed in parallel for multiple transformer areas to achieve batch photovoltaic capacity assessment.

[0067] It should be understood that Figure 2 The modules and references described in the document Figure 1 The steps described in the text correspond to those in the method described above. Therefore, the operations, features, and corresponding technical effects described above also apply to the method described in the text. Figure 2 The various modules in the document will not be described in detail here.

[0068] In other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor performs the photovoltaic capacity calculation method based on data-driven and neural networks in any of the above method embodiments. S1. Collect historical operating data of the transformer area and preprocess the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer area outlet voltage data and load rate data; S2. Extract features from the standardized data to construct a feature vector containing power features and timing features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. S3. Set safety and stability constraints for the operation of the distribution area. Based on the voltage evaluation model of the distribution area, iteratively simulate the photovoltaic grid connection scenario and perform stability verification to deduce the maximum photovoltaic capacity that the distribution area can open.

[0069] Computer-readable storage media may include a stored program area and a stored data area, wherein the stored program area may store an operating system and an application program required for at least one function; the stored data area may store data created based on the use of the data-driven and neural network-based photovoltaic capacity calculation system, etc. Furthermore, the computer-readable storage medium may include high-speed random access memory, and may also include memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include memory remotely disposed relative to a processor, which can be connected to the data-driven and neural network-based photovoltaic capacity calculation system via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0070] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present invention, such as... Figure 3 As shown, the device includes a processor 310 and a memory 320. The electronic device may also include an input device 330 and an output device 340. The processor 310, memory 320, input device 330, and output device 340 can be connected via a bus or other means. Figure 3 Taking a bus connection as an example, the memory 320 is the computer-readable storage medium described above. The processor 310 executes various server functions and data processing by running non-volatile software programs, instructions, and modules stored in the memory 320, thereby implementing the photovoltaic capacity calculation method based on data-driven and neural networks as described in the above method embodiment. The input device 330 can receive input digital or character information and generate key signal inputs related to user settings and function control of the photovoltaic capacity calculation system based on data-driven and neural networks. The output device 340 may include a display screen or other display device.

[0071] The aforementioned electronic device can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0072] In one implementation, the above-described electronic device is applied to a data-driven and neural network-based photovoltaic capacity calculation system for a client, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: S1. Collect historical operating data of the transformer area and preprocess the historical operating data to obtain standardized data; the historical operating data includes at least net load active power time series data, transformer area outlet voltage data and load rate data; S2. Extract features from the standardized data to construct a feature vector containing power features and timing features, and train a neural network model using the feature vector and the corresponding historical voltage data to obtain a data-driven transformer area voltage assessment model. S3. Set safety and stability constraints for the operation of the distribution area. Based on the voltage evaluation model of the distribution area, iteratively simulate the photovoltaic grid connection scenario and perform stability verification to deduce the maximum photovoltaic capacity that the distribution area can open.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data-driven and neural network-based photovoltaic capacity calculation method, characterized in that, The method comprises the following steps: S1, collecting historical operation data of a transformer area, and preprocessing the historical operation data to obtain standardized data; The historical operation data at least includes active power time series data of net load, outlet voltage data of the transformer area, and load rate data; S2, extracting features from the standardized data, constructing a feature vector containing power features and time series features, and training a neural network model using the feature vector and corresponding historical voltage data to obtain a data-driven transformer area voltage evaluation model; S3, setting safety and stability constraints for transformer area operation, based on the transformer area voltage evaluation model, iteratively simulating a photovoltaic grid-connected scenario and performing stability checking to back-propagate the maximum photovoltaic capacity that can be opened in the transformer area.

2. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The historical operation data of the transformer area is collected by a high-speed power line carrier communication (HPLC) intelligent total meter or concentrator installed on the low-voltage side of the transformer of the transformer area, the sampling interval is 15 minutes, and the data time span is not less than 1 year.

3. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The preprocessing of the historical operation data comprises the following steps: Missing data is filled by linear interpolation; For data segments with more than 4 consecutive missing points, the average curve of the same type of date is used for replacement; Abnormal values are identified based on the 3σ criterion, and the sliding average of adjacent normal data is used for replacement; The net load, voltage and load rate data are normalized to the interval [0, 1].

4. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The safety and stability constraints include: The outlet voltage deviation of the transformer area is not more than ±7% of the rated voltage, the load rate of the main line is not more than 80%, and the load rate of the distribution transformer is not more than 85%.

5. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The neural network model is an artificial neural network comprising one input layer, two hidden layers and one output layer; The number of neurons in the input layer is the same as the dimension of the feature vector; Each hidden layer contains 32 neurons, and the activation function uses the ReLU function. Because of the grid search test, among the 16, 32 and 64 neuron structures, the 32 layer has the lowest RMSE in the validation set and the optimal training efficiency; The output layer contains 1 neuron for outputting the transformer area voltage prediction value; The neural network model uses the Adam optimizer to minimize the root mean square error (RMSE) as the training target, and uses the early stopping method to prevent overfitting.

6. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The iterative simulation of the photovoltaic grid-connected scenario comprises the following steps: S31, setting an initial value of the photovoltaic capacity based on the number of transformer area users; S32, inputting the initial value into a photovoltaic power prediction model to obtain a photovoltaic output curve, and superimposing the photovoltaic output curve on the transformer area basic load curve to generate a simulated net load curve after grid connection; S33, inputting the simulated net load curve into the transformer area voltage evaluation model to obtain a voltage prediction value, and calculating the load rate of the line and transformer; S34, checking whether the voltage prediction value and the load rate meet the safety and stability constraints; If yes, increase the photovoltaic capacity and repeat steps S32-S34; If not, reduce the photovoltaic capacity and repeat steps S32-S34; Until the maximum photovoltaic capacity that meets all the constraints is found.

7. The data-driven and neural network-based photovoltaic capacity calculation method of claim 1, wherein, The method further comprises step S4: Modularize and standardize the data processing, model calling and capacity calculation process based on the big data cluster of the enterprise middle platform. The steps S1-S3 are executed in parallel for multiple transformer areas by using a distributed computing framework to achieve batch photovoltaic capacity evaluation.

8. A data-driven and neural network based photovoltaic capacity calculation system, characterized in that, The method comprises the following steps: a data acquisition and preprocessing module configured to acquire historical operation data of a transformer area and preprocess the historical operation data to obtain standardized data; The historical operation data at least includes active power time series data of net load, transformer area outlet voltage data and load rate data; a feature extraction and model training module configured to extract features from the standardized data, construct a feature vector containing power features and time series features, and train a neural network model using the feature vector and corresponding historical voltage data to obtain a data-driven transformer area voltage evaluation model; a capacity calculation and backstepping module configured to set safety and stability constraints for transformer area operation, iteratively simulate photovoltaic grid-connected scenarios and perform stability checking based on the transformer area voltage evaluation model to backstep the maximum photovoltaic capacity that can be opened in the transformer area; a big data platform interface module configured to modularize and standardize the data processing, model calling and capacity calculation process by relying on the big data cluster of the enterprise middle platform; and execute the steps S1-S3 in parallel for multiple transformer areas by using a distributed computing framework to achieve batch photovoltaic capacity evaluation.

9. An electronic device, comprising: The method comprises the following steps: at least one processor and a memory connected in communication with the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the method of any one of claims 1 to 7.

Images