A refined control method and system for new energy grid connection

Abstract

The present invention discloses a refined control method and system for new energy grid connection, which relates to the field of data encryption technology, including real-time collection of meteorological environment, power grid status and real-time power load data of users, and data preprocessing to form a real-time data warehouse; advanced feature construction and deep learning model analysis of the real-time data warehouse to form a special load association report; dynamic automatic adjustment of grid connection strategy, configuration of early warning rules, real-time tracking of grid connection performance and special load processing, output of real-time early warning information, and rapid feedback and adjustment. The present invention integrates multi-dimensional dynamic data streams and power grid operation history archives through an intelligent control architecture that integrates deep learning and advanced prediction algorithms, promotes efficient coordination of information across data sources, breaks the bottleneck of traditional static models in dealing with new energy uncertainty and power grid complexity, and realizes high-precision joint prediction of new energy power generation and power grid load.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy grid connection, and in particular to a method and system for fine-grained control of new energy grid connection. Background Art

[0002] With the transformation and upgrading of the global energy structure, the grid-connected utilization of new energy, especially renewable energy such as wind power and solar energy, has become an important way to promote green and low-carbon development. However, the intermittent and uncertain nature of new energy power generation has brought unprecedented challenges to the stable operation of the power grid. In order to achieve the efficient and safe integration of new energy into the traditional power grid system, the research and practice of new energy grid-connected technology is particularly important, among which refined management and control has become the key.

[0003] In the context of the rapid development of new energy grid-connected technology, although there are many technologies dedicated to improving the access and dispatching efficiency of renewable energy, existing technologies still face core challenges in the face of complex and changeable meteorological environments and user load demands. First, most systems focus on single-aspect optimization, such as separate load forecasting or simple grid-connected strategy design, and lack comprehensive solutions from a global perspective. Secondly, in the face of extreme climate events and sudden changes in user behavior patterns, existing models often find it difficult to accurately predict special loads and their impact on the grid's acceptance capacity in a timely and accurate manner, resulting in delayed and inefficient control strategies. At the same time, traditional linear programming methods are unable to cope with nonlinear, time-related new energy grid-connected problems, and are unable to accurately capture the dynamic characteristics of the system, affecting the stability and efficiency of the grid. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a refined control method and system for new energy grid connection to solve the problems of insufficient prediction accuracy, single control strategy and limited ability of the system to cope with complex changes in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a method for refined control of new energy grid connection, which includes real-time collection of meteorological environment, grid status and real-time user power load data, and data preprocessing to form a real-time data warehouse; high-level feature construction and deep learning model analysis of the real-time data warehouse to form a special load association report; based on the special load association report, special load probability prediction data is generated by using time series analysis and long short-term memory network models; combining the special load probability prediction data and grid status data, a grid acceptance capacity prediction curve is generated through a nonlinear programming algorithm; based on the long short-term memory network model and the grid acceptance capacity prediction curve, a dynamic automatic adjustment grid connection strategy is designed; the dynamic automatic adjustment grid connection strategy is implemented, warning rules are configured, grid connection performance and special load processing are tracked in real time, real-time warning information is output, and rapid feedback and adjustment are performed.

[0008] As a preferred solution of the refined control method of new energy grid connection described in the present invention, the data preprocessing includes error value detection, data calibration, data aggregation, normalization and missing value processing.

[0009] As a preferred solution of the refined control method for new energy grid connection of the present invention, the real-time data warehouse is subjected to advanced feature mapping and deep learning model analysis to form a special load association report, and the specific steps are as follows:

[0010] Based on the real-time data warehouse, the spectral clustering algorithm is used for feature extraction and Laplace matrix advanced feature mapping. The expression of advanced feature mapping is:

[0011] φ=eig(L) k

[0012] Among them, φ is the eigenvector matrix, eig is the process of eigenvalue decomposition, L is the Laplace matrix, and k is the first k eigenvalues ​​selected;

[0013] The feature vector matrix φ is input into the bidirectional recurrent neural network model, and the time series data is processed in combination with the attention weight to perform deep learning model analysis. The attention weight calculation expression is:

[0014] α t =softmax(v T tanh(P*φ⊙h t ++Qs (t-1) ))

[0015] Among them, α t is the attention weight of historical information at time step t, softmax is the activation function, v T is the transpose of the weight vector multiplied by the output of the hyperbolic tangent function, tanh is the hyperbolic tangent function, st-1 is the time step The context vector, W[φ⊙h t ] is the weight feature vector φ and the hidden state h t The weight matrix of the multiplied result, Q is the context vector s of the previous time step t-1 The weight matrix of

[0016] Using attention weight α t Traverse all past hidden states h i Perform weighted summation to form a context vector, context vector c t The expression is:

[0017]

[0018] Finally, using the context vector c t , through a fully connected layer to predict the special load value y t :

[0019] y t =Lr(c t )+b

[0020] Among them, y t is the output prediction result at time step t, Lr is a fully connected layer, and b is the bias term;

[0021] Based on the special load value y t , the gradient boosting tree loss model is used to perform feature importance ranking analysis to identify the key factors affecting special loads. The total loss model L is expressed as:

[0022]

[0023] Among them, L′(y t ,f(x i )) is the gradient boosting tree local loss model, x i is the i-th sample in the input data of the gradient boosting tree local loss model, f(x i ) is the sample prediction value, Ω(g) is the regularization term;

[0024] The feature importance ranking analysis of the gradient boosted tree loss function is combined with the prediction results of the deep learning model to compile a special load association report.

[0025] As a preferred solution of the refined control method for new energy grid connection of the present invention, based on the special load association report, the special load probability prediction data is generated by using time series analysis and long short-term memory network model, and the specific steps are as follows:

[0026] Summarize the special load historical data of actual load value, timestamp, weather conditions and holiday information in the special load association report, remove outliers, fill missing data, perform normalization preprocessing, and generate an original data feature set;

[0027] Apply time series analysis to pre-processed special load historical data to identify trend, seasonality and random component characteristics in load data;

[0028] The autoregressive integrated moving average model is used to extract the autocorrelation and partial autocorrelation features of the sequence, and the long short-term memory network model is used to assist in understanding the sequence structure.

[0029] The random component features obtained from time series analysis are combined with the original data feature set to form an enhanced feature set;

[0030] By using the long short-term memory network model and receiving the enhanced feature set, the output layer of the long short-term memory network model outputs the probability distribution of the load falling in different intervals at a specific future time point, and generates special load probability prediction data.

[0031] As a preferred solution of the refined control method for new energy grid connection of the present invention, the special load probability prediction data and the grid state data are combined to generate a grid acceptance capacity prediction curve through a nonlinear programming algorithm, and the specific steps are as follows:

[0032] The generated special load probability forecast data is spatially and temporally aligned with the real-time grid status data to form a comprehensive data set;

[0033] Using comprehensive data sets to build nonlinear programming models, the grid acceptance capacity simulation of load demand, power constraints and system stability is carried out for special load probability forecast data;

[0034] Generate multiple future load scenarios through Monte Carlo simulation, each scenario is used as input to the nonlinear programming model to obtain a scenario result set;

[0035] Statistically analyze all scenarios in the scenario result set, determine the upper and lower limits and probability distribution of the acceptance capacity, and generate a grid acceptance capacity prediction curve.

[0036] As a preferred solution of the refined control method of new energy grid connection of the present invention, based on the long short-term memory network model and the grid acceptance capacity prediction curve, a dynamic automatic adjustment grid connection strategy is designed, and the specific steps are as follows:

[0037] The special load probability prediction data output by the long short-term memory network model is synchronously fused with the data of the grid acceptance capacity prediction curve in time and space to form a fusion data set X f ;

[0038] By defining the interactive characteristics of the load probability and the power grid state in the window, the dynamic characteristics of the power grid in the current and future period of time can be reflected. int The expression is:

[0039]

[0040] Among them, X int (t) is the interactive characteristic value of the load probability and the power grid state in the window at time point t, w(Δt) is the weight function in the time window, P pb (t+Δt) is the load probability distribution at time point t+Δt, S gd (t) is the grid state feature set at time point t;

[0041] Based on the fusion feature set X f With interaction feature X int , design an adaptive adjustment strategy function F,

[0042]

[0043] Among them, argmax is the value of the independent variable corresponding to the maximum value of the function, Pin is the grid-connected input power, U is the benefit function, and C is the cost function;

[0044] A Q value update rule function is constructed to cope with the real-time changing grid conditions and adjust the grid-connected input power Pin in real time. The Q value update rule function expression is:

[0045]

[0046] Among them, Q new (s t ,a t ) is in state s at time step t t Take action when t The updated new value Q, (1-α) is the learning rate, Q(s t ,a t ) is in state s at time step t t Take action when t The old value of Q, r t+1 To take action t The reward immediately after , γ is the discount factor, is the state s at the next time step t+1 t+1 The maximum value Q among all actions is a′ in state s t+1 All next actions to consider;

[0047] According to the real-time adjustment of the grid-connected input power Pin, the early warning rules are configured to generate a dynamic automatic adjustment grid-connected strategy.

[0048] As a preferred solution of the refined control method for new energy grid connection of the present invention, the configuration of early warning rules comprises the following specific steps:

[0049] The configuration of the early warning rules requires setting thresholds and taking countermeasures based on early warning indicators, wherein the early warning indicators include grid-connected power deviation, grid frequency stability, voltage level, load forecast error and grid acceptance capacity utilization rate;

[0050] Setting and countermeasures of grid-connected power deviation threshold:

[0051] When the deviation between the actual grid-connected power and the expected power exceeds a threshold, the output power of the renewable energy power generation system is automatically adjusted, or the user's power demand is reduced through a demand response strategy.

[0052] As a preferred solution of the refined control method for new energy grid connection of the present invention, the grid frequency stability and voltage level, threshold setting and countermeasures in the early warning indicators are specifically performed as follows:

[0053] According to the impact of the change of input power Pin on the grid frequency, voltage level, load forecast error, and grid acceptance capacity utilization, the frequency deviation threshold is set to ±0.2Hz, the rated voltage of the voltage level is ±5%, the load forecast error threshold is ±2%, and the grid acceptance capacity utilization threshold is 80%;

[0054] Setting of grid frequency stability threshold and countermeasures:

[0055] When the frequency deviation exceeds the threshold of ±0.2Hz, the grid frequency is restored by adjusting the output of the generator in the automatic generation control system;

[0056] Voltage level threshold setting and countermeasures:

[0057] When the voltage level exceeds the threshold of ±5%, static VAR generator control is used to regulate the voltage.

[0058] As a preferred solution of the refined control method for new energy grid connection of the present invention, the load forecast error and grid acceptance capacity utilization rate in the early warning indicators, the setting of thresholds and the response measures are specifically as follows:

[0059] Load forecast error threshold setting and countermeasures:

[0060] When the forecast error exceeds the threshold of ±2%, the reserve capacity is adjusted or demand response resources are called to reduce the impact of the forecast error on the power grid;

[0061] Setting and countermeasures of grid acceptance capacity utilization threshold:

[0062] When the capacity utilization rate is equal to or exceeds the threshold of 80%, the operation status of the power grid is optimized through demand-side management or adjustment of power generation plan.

[0063] In a second aspect, the present invention provides a refined management and control system for new energy grid connection, including a data collection module, a feature construction and deep learning analysis module, a load prediction module, a power grid acceptance capacity prediction module, a dynamic automatic adjustment grid connection strategy design module and an early warning rule configuration module; the data collection module is used to collect meteorological environment, power grid status and user real-time power load data in real time, and perform data preprocessing to form a real-time data warehouse; the feature construction and deep learning analysis module is used to perform advanced feature construction and deep learning model analysis on the real-time data warehouse to form a special load association report; the load prediction module is used to generate special load probability prediction data based on the special load association report by using time series analysis and long short-term memory network model; the power grid acceptance capacity prediction module is used to combine the special load probability prediction data and power grid status data to generate a power grid acceptance capacity prediction curve through a nonlinear programming algorithm; the dynamic automatic adjustment grid connection strategy design module is used to design a dynamic automatic adjustment grid connection strategy based on the long short-term memory network model and the power grid acceptance capacity prediction curve; the early warning rule configuration module is used to implement the dynamic automatic adjustment grid connection strategy, configure early warning rules, track the grid connection performance and special load processing in real time, output real-time early warning information, and perform rapid feedback and adjustment.

[0064] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, any step of the method for refined control of new energy grid connection as described in the first aspect of the present invention is implemented.

[0065] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the method for refined control of new energy grid connection as described in the first aspect of the present invention is implemented.

[0066] The beneficial effects of the present invention are as follows: the present invention realizes comprehensive optimization of the new energy grid connection process by integrating advanced data processing technology, deep learning model and nonlinear programming algorithm; collects and integrates meteorological environment, power grid status and user real-time power load data in real time, forms a real-time data warehouse through data preprocessing, and then uses advanced feature construction methods such as spectral clustering and bidirectional recurrent neural network to deeply explore the intrinsic connection between special load and system status. On this basis, the long short-term memory network model and time series analysis are introduced to accurately predict the probability distribution of special load, overcoming the prediction limitations of traditional methods in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0068] Figure 1 This is a flow chart of the refined control method for new energy grid connection in Example 1. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0070] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0071] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0072] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and provides a method for fine-grained control of new energy grid connection, comprising the following steps:

[0073] S1. Collect meteorological environment, power grid status and user real-time power load data in real time, and perform data preprocessing to form a real-time data warehouse.

[0074] Furthermore, the system continuously and frequently collects three types of data through sensor networks and smart meters deployed at key locations, as well as data interfaces with meteorological departments: meteorological environmental parameters such as wind speed, light intensity, temperature, humidity, etc., grid status information including but not limited to voltage level, current intensity, frequency stability, line load status, etc., and real-time user power load data reflecting the power consumption patterns of different users in different time periods. These data constitute the basic raw materials for analyzing the status of new energy grid connection;

[0075] Going further, we enter the data preprocessing stage, which includes five key steps to purify and standardize the data and eliminate noise and errors that may affect the quality of analysis. Data preprocessing includes:

[0076] Error value detection, using statistical methods or machine learning algorithms to identify outliers in the data, such as misreadings caused by equipment failure or communication interference, to ensure the authenticity and accuracy of the data;

[0077] Data calibration: Correct the measured values ​​according to known standards or historical data to eliminate systematic deviations caused by sensor drift or environmental factors, and improve data consistency and comparability;

[0078] Data aggregation: integrating data scattered at different time granularities or geographical locations as needed to form a unified time series or spatial distribution data set to facilitate subsequent comprehensive analysis;

[0079] Normalization is the process of mapping values ​​of different scales or ranges to the same interval, usually 0 to 1, through linear transformation, so that data from different sources can be fairly compared under the same framework and avoid analytical bias caused by different dimensions;

[0080] Missing value processing uses interpolation, model prediction or other statistical means to fill in blank values ​​in the data set caused by various reasons, maintain the continuity and integrity of the data, and reduce information loss caused by missing values.

[0081] S2. Perform advanced feature construction and deep learning model analysis on the real-time data warehouse to form special load correlation reports.

[0082] Furthermore, based on the real-time data warehouse, the spectral clustering algorithm is used for feature extraction and Laplacian matrix advanced feature mapping to capture the topological structure of the data. The advanced feature mapping expression is:

[0083] φ=eig(L) k

[0084] Among them, φ is the eigenvector matrix, eig is the process of eigenvalue decomposition, L is the Laplace matrix, and k is the first k eigenvalues ​​selected;

[0085] It should be noted that

[0086] The feature vector matrix φ is input into the bidirectional recurrent neural network model, and the time series data is processed in combination with the attention weight to perform deep learning model analysis to capture the long-term dependencies in the data sequence. The attention weight calculation expression is:

[0087] α t =softmax(v Ttanh(P*φ⊙h t ++Qs (t-1) ))

[0088] Among them, α t is the attention weight of historical information at time step t, softmax is the activation function, v T is the transpose of the weight vector multiplied by the output of the hyperbolic tangent function, tanh is the hyperbolic tangent function, s t-1 is the context vector at time step t-1, W[φ⊙h t ] is the weight feature vector φ and the hidden state h t The weight matrix of the multiplied result, Q is the context vector s of the previous time step t-1 The weight matrix of

[0089] It should be noted that the softmax function is an activation function widely used in the fields of machine learning and deep learning, and plays a core role in multi-classification problems. The following is the specific definition and form of the softmax function:

[0090] The softmax function receives a real number vector z as input. The input real number vector z is expressed as:

[0091] z=[z1,z2,…,z K ]

[0092] i=1,2,...,K

[0093] Among them, [z1,z2,…,z K ] is the unnormalized score of the model for the i-th category;

[0094] The output of the softmax function is also a vector, and the output vector σ is expressed as:

[0095] σ(z)=[σ(z1),σ(z2),…,σ(z K )]

[0096] i=1,2,...,K

[0097] Among them, σ(z i )For each i=1,2,...,K, the softmax formula is followed;

[0098] The softmax formula is expressed as:

[0099]

[0100] Among them, σ(z) iis actually the ratio of the score of the i-th category to the total score of all categories, indicating the unnormalized score of the model for each classification, z j Score all categories, z i is the score of the i-th category, K is the total number of categories, that is, the length of the real number vector Z.

[0101] Furthermore, using the attention weight α t Traverse all past hidden states h i Perform weighted summation to form a context vector, which contains the importance of historical information. The context vector c t The expression is:

[0102]

[0103] Finally, using the context vector c t , through a fully connected layer to predict the special load value y t :

[0104] y t =Lr(c t )+b

[0105] Among them, y t is the output prediction result at time step t, Lr is a fully connected layer, and b is the bias term;

[0106] Based on the special load value y t , the gradient boosting tree loss model is used to perform feature importance ranking analysis and identify the key factors affecting special loads. The gradient boosting tree loss model iteratively constructs multiple decision trees and focuses on the samples with poor predictions in the previous round in each round, gradually reducing the prediction error. The total loss model L is expressed as:

[0107]

[0108] Among them, L′(y t ,f(x i )) is the gradient boosting tree local loss model, x i is the i-th sample in the input data of the gradient boosting tree local loss model, f(x i ) is the sample prediction value, Ω(g) is the regularization term to prevent overfitting;

[0109] The feature importance ranking analysis of the gradient boosted tree loss function is combined with the prediction results of the deep learning model to deeply analyze which features are most critical for special load prediction. These analysis results are integrated into a detailed report, namely the special load association report.

[0110] S3. Based on the special load association report, the special load probability forecast data is generated using time series analysis and long short-term memory network model.

[0111] Furthermore, the historical data of special loads including the actual load value, timestamp, weather conditions such as temperature, humidity, wind speed, and whether it is a holiday and other key information in the special load association report are summarized, and Z-score outliers are removed, missing data are filled by linear interpolation, and normalization preprocessing is performed to eliminate the impact of dimensions, improve the efficiency of model training, make it fall within a unified scale range, and generate the original data feature set as the basic feature set of model input;

[0112] Apply time series analysis to pre-processed special load historical data to identify the trend, seasonality and random component characteristics in the load data, so as to facilitate the understanding of the inherent laws of load changes;

[0113] The autoregressive integrated moving average model is used to extract the autocorrelation and partial autocorrelation features of the sequence, and the long short-term memory network model is used to assist in understanding the sequence structure.

[0114] The random component features obtained from time series analysis are combined with the original data feature set to form an enhanced feature set. This step enhances the richness of the model input and helps improve the accuracy of the prediction.

[0115] By using the long short-term memory network model and receiving the enhanced feature set, the output layer of the long short-term memory network model outputs the probability distribution of the load falling in different intervals at a specific future time point. This means that the model not only predicts the load value, but also provides a probabilistic framework to evaluate the uncertainty of the prediction, which is particularly important for decision support. It generates special load probability prediction data, enabling decision makers to better formulate response strategies, such as scheduling resources and preventing overloads.

[0116] S4. Combining the special load probability prediction data and the grid status data, a grid acceptance capacity prediction curve is generated through a nonlinear programming algorithm.

[0117] Furthermore, the generated special load probability forecast data is spatially and temporally aligned with the real-time grid status data to ensure the consistency of the two data in time and space dimensions, so that the load forecast data can accurately reflect the actual operation status of the grid at a specific time point and location, thus forming a comprehensive data set;

[0118] It should be noted that the generated special load probability prediction data are spatially and temporally aligned with the real-time grid status data, which improves the accuracy and applicability of the model input data and provides a unified and accurate data platform for subsequent nonlinear programming models.

[0119] Using the comprehensive data set to construct a nonlinear programming model, the load demand is given by the probability of different load levels given by the special load probability forecast, and the power constraints such as transmission line capacity limit, generator output range and system stability are simulated for the grid acceptance capacity;

[0120] It should also be noted that the operation status of the power grid is simulated when facing different load demands, and its ability to accommodate additional special loads is evaluated, while ensuring the safety and stability of the system operation.

[0121] Through Monte Carlo, different load levels in the special load probability forecast data are randomly selected to simulate and generate multiple future load scenarios, such as the fluctuation of renewable energy output and the randomness of user behavior. Each scenario is used as the input of the nonlinear programming model, which increases the comprehensiveness and robustness of the analysis, enables the model to cope with uncertainty, provides a richer perspective for decision-making, and finally obtains a set of scenario results;

[0122] Statistically analyze all scenarios in the scenario result set, determine the upper and lower limits of the acceptance capacity and the probability distribution, and generate a grid acceptance capacity prediction curve to intuitively display the maximum and minimum acceptance load levels of the grid under different probabilities, as well as the distribution of various possible states in the middle.

[0123] It should be noted that the grid acceptance capacity prediction curve provides a clear visual reference for grid operators, helping them understand the grid acceptance limit under specific conditions, formulate more reasonable dispatching strategies, optimize resource allocation, ensure safe and efficient operation of the grid, and provide a scientific basis for future grid planning.

[0124] S5. Based on the long short-term memory network model and the grid acceptance capacity prediction curve, a dynamic automatic adjustment grid connection strategy is designed.

[0125] Furthermore, the special load probability prediction data output by the long short-term memory network model is fused with the data of the grid acceptance capacity prediction curve in time and space to form a fusion data set X f ,The ,synchronous integration of time and space ,provides a unified view for subsequent strategy design that comprehensively ,considers the uncertainty of load forecasting and the current and future ,acceptance capacity of the power grid;

[0126] By defining the interactive characteristics of the load probability and the power grid state within the window, the dynamic characteristics of the power grid in the current and future period of time can be reflected, providing a basis for formulating more accurate grid connection strategies. int The expression is:

[0127]

[0128] Among them, X int(t) is the interactive characteristic value of the load probability and the power grid state in the window at time point t, w(Δt) is the weight function in the time window, which is used to adjust the importance of data at different time points.

[0129] P pb (t+Δt) is the load probability distribution at time point t+Δt, S gd (t) is the grid state feature set at time point t;

[0130] Based on the fusion feature set X f With interaction feature X int , design an adaptive adjustment strategy function F,

[0131]

[0132] Among them, argmax is the value of the independent variable corresponding to the maximum value of the function, Pin is the grid-connected input power, U is the benefit function, and C is the cost function;

[0133] A Q value update rule function is constructed to cope with the real-time changing grid conditions and adjust the grid-connected input power Pin in real time. The Q value update rule function expression is:

[0134]

[0135] Among them, Q new (s t ,a t ) is in state s at time step t t Take action when t The updated new value Q, (1-α) is the learning rate, Q(s t ,a t ) is in state s at time step t t Take action when t The old value of Q, r t+1 To take action t The reward immediately after , γ is the discount factor, is the state s at the next time step t+1 t+1 The maximum value Q among all actions is a′ in state s t+1 All next actions to consider;

[0136] According to the real-time adjustment of the grid-connected input power Pin, the early warning rules are configured to generate a dynamic automatic adjustment grid-connected strategy, which can effectively deal with the randomness and uncertainty of the grid load, ensure the stability and economy of the grid operation, and at the same time improve the grid-connected efficiency of renewable energy and reduce unnecessary energy waste and system pressure.

[0137] S6. Implement dynamic automatic adjustment of grid-connected strategies, configure early warning rules, track grid-connected performance and special load processing in real time, output real-time early warning information, and conduct rapid feedback and adjustments.

[0138] Furthermore, configuring early warning rules requires setting thresholds and taking countermeasures based on early warning indicators, including grid-connected power deviation, grid frequency stability, voltage level, load forecast error, and grid acceptance capacity utilization;

[0139] Setting and countermeasures of grid-connected power deviation threshold:

[0140] When the deviation between the actual grid-connected power and the expected power exceeds a threshold, the output power of the renewable energy power generation system is automatically adjusted, or the user's power demand is reduced through a demand response strategy.

[0141] Setting of grid frequency stability threshold and countermeasures:

[0142] When the frequency deviation exceeds the threshold of ±0.2Hz, the grid frequency is restored by adjusting the output of the generator in the automatic generation control system;

[0143] Voltage level threshold setting and countermeasures:

[0144] When the voltage level exceeds the threshold of ±5%, static VAR generator control is used to regulate the voltage.

[0145] Load forecast error threshold setting and countermeasures:

[0146] When the forecast error exceeds the threshold of ±2%, the reserve capacity is adjusted or demand response resources are called to reduce the impact of the forecast error on the power grid;

[0147] Setting and countermeasures of grid acceptance capacity utilization threshold:

[0148] When the capacity utilization rate is equal to or exceeds the threshold of 80%, the operation status of the power grid is optimized through demand-side management or adjustment of power generation plan.

[0149] It should be noted that the configuration of early warning rules is used to monitor the operation status of the power grid in real time, especially the key indicators related to the connection of new energy to the grid, to ensure that the system can identify potential problems in a timely manner and respond quickly, thereby avoiding system imbalance or failure. By configuring reasonable early warning rules, early warning can be given, prompting the system to take preset adjustment measures to maintain the safe and stable operation of the power grid.

[0150] It should also be noted that these early warning rules and response measures together constitute a closed-loop control system that can not only respond to changes in the state of the power grid in real time, but also take proactive measures to prevent possible problems. Through refined management, not only the flexibility and reliability of power grid operation are improved, but also the effective grid connection and utilization of renewable energy are promoted.

[0151] This embodiment also provides a file encryption system, including: a data collection module, a feature construction and deep learning analysis module, a load prediction module, a power grid acceptance capacity prediction module, a dynamic automatic adjustment grid-connected strategy design module and an early warning rule configuration module; the data collection module is used to collect meteorological environment, power grid status and user real-time power load data in real time, and perform data preprocessing to form a real-time data warehouse; the feature construction and deep learning analysis module is used to perform advanced feature construction and deep learning model analysis on the real-time data warehouse to form a special load association report; the load prediction module is used to generate special load probability prediction data based on the special load association report by using time series analysis and long short-term memory network model; the power grid acceptance capacity prediction module is used to combine special load probability prediction data and power grid status data, and generate a power grid acceptance capacity prediction curve through a nonlinear programming algorithm; the dynamic automatic adjustment grid-connected strategy design module is used to design a dynamic automatic adjustment grid-connected strategy based on the long short-term memory network model and the power grid acceptance capacity prediction curve; the early warning rule configuration module is used to implement the dynamic automatic adjustment grid-connected strategy, configure early warning rules, track the grid performance and special load processing in real time, output real-time early warning information, and perform rapid feedback and adjustment.

[0152] This embodiment also provides a computer device, which is suitable for the case of a refined control method for new energy grid connection, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the refined control method for new energy grid connection as proposed in the above embodiment.

[0153] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0154] The present embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for realizing refined control of new energy grid connection proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, disk or optical disk.

[0155] In summary, the present invention has created a set of efficient new energy grid-connected management system by combining real-time data collection, deep learning analysis, advanced prediction models and intelligent control strategies. It has greatly improved the grid-connected efficiency and grid stability, showing its important application value and practical significance in the integration of new energy.

[0156] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of the refined control method for new energy grid connection are provided.

[0157] A data collection network integrating meteorological stations, smart meters and power grid monitoring systems is established to ensure real-time acquisition of meteorological environmental parameters (such as temperature and wind speed), power grid status (voltage, current, frequency) and real-time power load data of users. The data preprocessing stage includes error value detection (using the triple standard deviation method to eliminate outliers), data calibration (linear regression correction based on historical data), data aggregation (aggregating data at a frequency of once every 15 minutes), normalization (minimum-maximum standardization to the interval [0,1]) and missing value processing (using a combination of forward filling and linear interpolation).

[0158] Firstly, the spectral clustering algorithm is used to extract features from the preprocessed real-time data warehouse, and the first five eigenvalues ​​are selected for Laplacian matrix advanced feature mapping.

[0159] Secondly, the obtained feature vector matrix is ​​input into the bidirectional recurrent neural network model, and combined with the attention mechanism (activated by tanh function), deep learning analysis is performed on the time series data. The dynamic adjustment of attention weights enables the model to more effectively capture the key information in the time series and then predict the special load value.

[0160] Finally, based on the gradient boosting tree loss model, the key factors affecting special loads are identified and a special load association report is formed. In order to intuitively show the difference between the invention content and the prior art, a set of comparative experiments is designed. The following four key parameters are selected for quantitative evaluation: load forecast accuracy, grid acceptance capacity forecast accuracy, warning response time and grid connection efficiency.

[0161] The details are shown in the following table:

[0162] Table 1 Experimental record table

[0163]

[0164]

[0165] Through the analysis of the data in the above table, it can be clearly seen that the effectiveness and superiority of the method of the present invention. Using the refined control method of new energy grid connection of the invention content, the load forecasting accuracy has increased by 7% to 92%, thanks to the efficient combination of deep learning model and attention mechanism, which can capture special load patterns more accurately. The prediction accuracy of the grid acceptance capacity has also been significantly improved to 90%, an increase of 12 percentage points over the prior art, which is attributed to the combination of nonlinear programming algorithm and Monte Carlo simulation, which effectively handles the prediction uncertainty under complex load scenarios. In addition, the warning response time was shortened from 120 seconds to 60 seconds, a reduction of 50%, which shows that the dynamic automatic adjustment of the grid connection strategy and the warning rule configuration module can quickly respond to changes in the grid state and take timely measures. The grid connection efficiency has been improved to 85%, an increase of 15% compared with the traditional method, proving that this method has obvious advantages in optimizing resource allocation and balancing supply and demand.

[0166] Through the above detailed experimental preparation, data processing, model construction and effect analysis results, the practicality of the refined control method of new energy grid connection is fully demonstrated. In terms of processing real-time data, predicting special loads, optimizing grid acceptance capacity prediction, and designing dynamic adjustment strategies, the invention content has demonstrated performance that exceeds the existing technology. In particular, it provides strong empirical data support for improving prediction accuracy, accelerating response speed, and improving grid connection efficiency, bringing significant progress to the field of new energy grid connection control.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A refined control method for new energy grid connection, characterized by: include, Collect meteorological environment, power grid status and user real-time power load data in real time, and perform data preprocessing to form a real-time data warehouse; Performing advanced feature mapping and deep learning model analysis on the real-time data warehouse to form a special load association report; Based on the special load association report, using time series analysis and long short-term memory network model, special load probability forecast data is generated; Combining the special load probability prediction data and the power grid state data, generating a power grid acceptance capacity prediction curve through a nonlinear programming algorithm; Based on the long short-term memory network model and the grid acceptance capacity prediction curve, a dynamic automatic adjustment grid connection strategy is constructed; Based on dynamic automatic adjustment of grid-connected strategies, configure early warning rules, track grid-connected performance and special load processing in real time, output real-time early warning information, and conduct rapid feedback and adjustments; Based on the special load association report, special load probability forecast data is generated using time series analysis and long short-term memory network model. The specific steps are as follows: Summarize the special load historical data of actual load value, timestamp, weather conditions and holiday information in the special load association report, remove abnormal values, fill missing data, perform normalization preprocessing, and generate an original data feature set; Apply time series analysis to pre-processed special load historical data to identify trend, seasonality and random component characteristics in load data; The autoregressive integrated moving average model is used to extract the autocorrelation and partial autocorrelation features of the sequence, and the long short-term memory network model is used to assist in understanding the sequence structure. The random component features obtained from time series analysis are combined with the original data feature set to form an enhanced feature set; The enhanced feature set is received by the long short-term memory network model, and the output layer of the long short-term memory network model outputs the probability distribution of the load falling in different intervals at a specific future time point to generate special load probability prediction data; Combining the special load probability prediction data and the power grid state data, a power grid acceptance capacity prediction curve is generated through a nonlinear programming algorithm. The specific steps are as follows: The generated special load probability forecast data is spatially and temporally aligned with the real-time grid status data to form a comprehensive data set; Using comprehensive data sets to build nonlinear programming models, the grid acceptance capacity simulation of load demand, power constraints and system stability is carried out for special load probability forecast data; Generate multiple future load scenarios through Monte Carlo simulation, each scenario is used as input to the nonlinear programming model to obtain a scenario result set; Statistically analyze all scenarios in the scenario result set, determine the upper and lower limits of the acceptance capacity and the probability distribution, and generate a grid acceptance capacity prediction curve; Based on the long short-term memory network model and the grid acceptance capacity prediction curve, a dynamic automatic adjustment grid connection strategy is designed. The specific steps are as follows: The special load probability prediction data output by the long short-term memory network model is synchronously fused with the data of the grid acceptance capacity prediction curve in time and space to form a fusion data set X f ; By defining the interactive characteristics of the load probability and the power grid state in the window, the dynamic characteristics of the power grid in the current and future period of time can be reflected. int The expression is: Among them, X int (t) is the interactive characteristic value of the load probability and the power grid state in the window at time point t, w(Δt) is the weight function in the time window, P pb (t+Δt) is the load probability distribution at time point t+Δt, S gd (t) is the grid state feature set at time point t; Based on the fusion feature set X f With interaction feature X int , design an adaptive adjustment strategy function F, Among them, arg max is the value of the independent variable corresponding to the maximum value of the function, Pin is the grid-connected input power, U is the benefit function, and C is the cost function; A Q value update rule function is constructed to cope with the real-time changing grid conditions and adjust the grid-connected input power Pin in real time. The Q value update rule function expression is: Among them, Q new (s t ,a t ) is in state s at time step t t Take action when t The updated new value Q, (1-α) is the learning rate, Q(s t ,a t ) is in state s at time step t t Take action when t The old value of Q, r t+1 To take action t The reward immediately after , γ is the discount factor, is the state s at the next time step t+1 t+1 The maximum value Q among all actions is a' in state s t+1 All next actions to consider; According to the real-time adjustment of the grid-connected input power Pin, the early warning rules are configured to generate a dynamic automatic adjustment grid-connected strategy.

2. The refined control method for new energy grid connection according to claim 1, characterized in that: The data preprocessing includes error value detection, data calibration, data aggregation, normalization and missing value processing.

3. The refined control method for new energy grid connection according to claim 2, characterized in that: Performing advanced feature mapping and deep learning model analysis on the real-time data warehouse to form a special load association report, the specific steps are: Based on the real-time data warehouse, the spectral clustering algorithm is used for feature extraction and Laplace matrix advanced feature mapping. The expression of advanced feature mapping is: φ=eg(L) k Among them, φ is the eigenvector matrix, eig is the process of eigenvalue decomposition, L is the Laplace matrix, and k is the first k eigenvalues ​​selected; The feature vector matrix φ is input into the bidirectional recurrent neural network model, and the time series data is processed in combination with the attention weight to perform deep learning model analysis. The attention weight calculation expression is: α t =softmax(v T tan h(P[φ⊙h t ]+Qs (t-1) )) Among them, α t is the attention weight of historical information at time step t, softmax is the activation function, v T is the transpose of the weight vector multiplied by the output of the hyperbolic tangent function, tan h is the hyperbolic tangent function, s t-1 is the context vector at time step t-1, P[φ⊙h t ] is the weight feature vector φ and the hidden state h t The weight matrix of the multiplied result, Q is the context vector s of the previous time step t-1 The weight matrix of Using attention weight α t Traverse all past hidden states h i Perform weighted summation to form a context vector, context vector c t The expression is: Finally, using the context vector c t , through a fully connected layer to predict the special load value y t : y t =Lr(c t )+b Among them, y t is the output prediction result at time step t, Lr is a fully connected layer, and b is the bias term; Based on the special load value y t , the gradient boosting tree loss model is used to perform feature importance ranking analysis to identify the key factors affecting special loads. The total loss model L is expressed as: Among them, L'(y t ,f(x i )) is the local loss function of the gradient boosting tree, x i is the i-th sample in the input data of the local loss function of the gradient boosting tree, f(x i ) is the sample prediction value, Ω(g) is the regularization term; The feature importance ranking analysis of the gradient boosting tree local loss function is combined with the prediction results of the deep learning model to compile a special load association report.

4. The refined control method for new energy grid connection according to claim 1, characterized in that: The specific steps of configuring the warning rules are as follows: The configuration of the warning rules requires setting thresholds and generating countermeasures based on warning indicators, wherein the warning indicators include grid-connected power deviation, grid frequency stability, voltage level, load forecast error and grid acceptance capacity utilization rate; Generate countermeasures through grid-connected power deviation threshold: When the deviation between the actual grid-connected power and the expected power exceeds a threshold, the output power of the renewable energy power generation system is automatically adjusted, or the user's power demand is reduced through a demand response strategy.

5. The refined control method for new energy grid connection according to claim 4, characterized in that: The grid frequency stability and voltage level in the early warning indicators, the setting of thresholds and the response measures are as follows: According to the impact of the change of input power Pin on the grid frequency, voltage level, load forecast error, and grid acceptance capacity utilization, the frequency deviation threshold is set to ±0.2Hz, the rated voltage of the voltage level is ±5%, the load forecast error threshold is ±2%, and the grid acceptance capacity utilization threshold is 80%; Setting of grid frequency stability threshold and countermeasures: When the frequency deviation exceeds the threshold of ±0.2Hz, the grid frequency is restored by adjusting the output of the generator in the automatic generation control system; Voltage level threshold setting and countermeasures: When the voltage level exceeds the threshold of ±5%, static VAR generator control is used to regulate the voltage.

6. The refined control method for new energy grid connection according to claim 5, characterized in that: The load forecast error and grid acceptance capacity utilization rate in the early warning indicators, the setting of thresholds and the response measures are as follows: Load forecast error threshold setting and countermeasures: When the forecast error exceeds the threshold of ±2%, the reserve capacity is adjusted or demand response resources are called to reduce the impact of the forecast error on the power grid; Setting and countermeasures of grid acceptance capacity utilization threshold: When the capacity utilization rate is equal to or exceeds the threshold of 80%, the operation status of the power grid is optimized through demand-side management or adjustment of power generation plan.

7. A refined control system for new energy grid connection, based on the refined control method for new energy grid connection according to any one of claims 1 to 6, characterized in that: It includes data collection module, feature construction and deep learning analysis module, load forecasting module, grid acceptance capacity forecasting module, dynamic automatic adjustment grid connection strategy design module and early warning rule configuration module; The data collection module is used to collect meteorological environment, power grid status and user real-time power load data in real time, and perform data preprocessing to form a real-time data warehouse; The feature construction and deep learning analysis module is used to perform advanced feature construction and deep learning model analysis on the real-time data warehouse to form a special load association report; The load forecasting module is used to generate special load probability forecasting data based on the special load association report by using time series analysis and long short-term memory network model; The grid acceptance capacity prediction module is used to combine the special load probability prediction data and the grid state data to generate a grid acceptance capacity prediction curve through a nonlinear programming algorithm; The dynamic automatic adjustment grid-connected strategy design module is used to design a dynamic automatic adjustment grid-connected strategy based on the long short-term memory network model and the grid acceptance capacity prediction curve; The warning rule configuration module is used to implement dynamic automatic adjustment of grid-connected strategies, configure warning rules, track grid-connected performance and special load processing in real time, output real-time warning information, and perform rapid feedback and adjustment.

Images