An algorithm for load output based on automatic adjustment of renewable energy output

Abstract

The present invention relates to the technical field of electric energy storage dispatching, and the present invention provides an algorithm for load output based on automatic adjustment of new energy output. The present invention aims to solve the problem of prediction and control of new energy output caused by dynamic changes in parameters of the power system. By collecting real-time, historical and predicted data of wind power and photovoltaics, combined with environmental and equipment status information, a new energy output prediction model is established, and the covariance matrix is ​​used to analyze the time correlation of wind power and photovoltaic output. The prediction accuracy is improved through cluster analysis and model optimization. The present invention matches the adjustment strategy in the knowledge base to achieve automatic adjustment of load output, improve the supply and demand balance and operational stability of the power system, and improve the utilization rate of new energy and the overall efficiency of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy storage dispatching, and in particular to an algorithm for automatically adjusting load output based on new energy output. Background Art

[0002] With the rapid development of new energy technologies, especially the widespread application of solar and wind energy around the world, the load output algorithm based on the automatic adjustment of output in the field of new energy has also made significant progress. The load output algorithm based on the automatic adjustment of output in the field of new energy aims to maximize the utilization of new energy output while ensuring the stability and reliability of the power system. At present, optimization algorithms based on mathematical modeling, artificial intelligence algorithms, and optimization strategies combined with market mechanisms can achieve automatic adjustment of new energy output and reasonable distribution of loads to varying degrees.

[0003] Renewable energy output data specifically refers to the power data generated by renewable energy generators in the power system, reflecting the power generation capacity and output of renewable energy solar energy and wind energy within a certain period of time. The load output of renewable energy output automatic adjustment refers to the automatic adjustment of the output of other loads in the power system according to the actual output of renewable energy solar energy and wind energy, so as to maintain the supply and demand balance and stable operation of the power system. The adjustment mechanism is an important part of the modern smart grid.

[0004] Although the existing technology has achieved remarkable results in the automatic adjustment of renewable energy output and load output, the following technical pain points still exist: the power system is composed of many components such as power generation equipment, transmission lines, substations, and distribution networks. These parts are interrelated to form a large and complex network. Any failure in any part of the power system may affect the normal operation of the entire system. The operation of the power system involves dynamic changes in multiple parameters such as voltage, current, and power. The parameters influence each other, making the automatic adjustment of the renewable energy output of the power system inaccurate. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an algorithm for automatically adjusting the load output based on the output of renewable energy, so as to solve the problem that the operation of the power system involves dynamic changes in multiple parameters such as voltage, current, power, etc., and the parameters influence each other, resulting in inaccurate automatic adjustment of the output of renewable energy in the power system.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows:

[0007] The present invention provides an algorithm for automatically adjusting load output based on new energy output, including:

[0008] Step S101, obtaining new energy output data and new energy output automatic adjustment demand data, the new energy output data includes wind power output data and photovoltaic output data, the wind power output data includes wind power real-time output data, wind power historical output data, wind power predicted output data, wind power environmental data, wind turbine status data and wind farm basic information, photovoltaic output data includes photovoltaic real-time output data, photovoltaic historical output data, photovoltaic predicted output data, photovoltaic panel performance data, photovoltaic environmental data, photovoltaic power station basic information, and establishing a new energy output prediction model based on the new energy output data;

[0009] Step S102, constructing a covariance matrix model of wind power and photovoltaic output, substituting the new energy output data into the covariance matrix model of wind power and photovoltaic output, and obtaining the time correlation information of wind power and photovoltaic output;

[0010] Step S103, clustering analysis is performed on the prediction data obtained by the new energy output prediction error model and the time correlation of wind power and photovoltaic output, to obtain the clustering result of the new energy output data, labeling and sorting the clustering result of the new energy output data, and verifying the data corresponding to each label in turn. During the verification process, a corresponding relationship is established between each label and the new energy output prediction model, and valid data and invalid data are screened out;

[0011] Step S104, optimizing the new energy output prediction model corresponding to the invalid data to obtain an optimized new energy output prediction model;

[0012] Step S105, obtaining real-time new energy output data, substituting the real-time new energy output data into the optimized new energy output prediction model, matching in the knowledge base according to the result of the optimization algorithm, and obtaining the load output adjustment strategy.

[0013] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention includes:

[0014] Preprocessing the new energy output data to obtain preprocessed new energy output data;

[0015] receiving the influencing characteristics of the new energy output forecast and the new energy output forecast demand, the influencing characteristics of the new energy output forecast including time, weather conditions and equipment status;

[0016] Matching machine learning models in a preset model library according to the forecast demand for new energy output. The machine learning models include time series analysis, regression models, integrated learning methods or deep learning models;

[0017] Using historical data as a training set, the influencing features of renewable energy output prediction are input into the matched machine learning model, and model training is performed to obtain a renewable energy output prediction model.

[0018] Deploy the trained new energy output prediction model to the server, receive new new energy output data as input data in real time, and output the prediction results of new energy output;

[0019] Using the newly acquired new energy output data according to a preset time period to construct a new energy output prediction model to obtain an updated new energy output prediction model;

[0020] A simulation mirror data processing area is established, and the new energy output data is copied to the simulation mirror data processing area. The new energy output data in the simulation mirror data processing area is predicted by the updated new energy output prediction model to obtain the prediction result of the updated new energy output prediction model. The prediction result of the updated new energy output prediction model is compared with the actual output data in the new energy output data. If the actual output data in the new energy output data is consistent with the prediction result of the updated new energy output prediction model, the updated new energy output prediction model replaces the existing new energy output prediction model.

[0021] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S102, includes:

[0022] Extract wind power and photovoltaic output time series data rows from the database, determine the time series of wind power and photovoltaic output, each time series corresponds to a variable, the variables include wind power output and photovoltaic output, determine the number of observations and time intervals of each variable, so that the length of each variable time series and the sampling rate of each variable are consistent;

[0023] Initialize an empty matrix, the dimension of the initialized matrix is ​​equal to the length of the time series multiplied by the number of variables;

[0024] Calculate the covariance between wind power and photovoltaic output at the time point corresponding to each element in the matrix to obtain the covariance calculation result between wind power and photovoltaic output;

[0025] Analyze the elements in the covariance calculation results between wind power and photovoltaic output, identify the positive correlation, negative correlation or irrelevance between wind power and photovoltaic output, and judge the strength of the correlation according to the size of the elements in the covariance calculation results between wind power and photovoltaic output.

[0026] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S102, includes:

[0027] Each element in the covariance matrix represents the covariance of wind power and photovoltaic output at a certain point in time. The size of the covariance value is used to reflect the strength of the correlation between wind power and photovoltaic output. The larger the absolute value of the covariance value, the stronger the correlation between wind power and photovoltaic output.

[0028] Determine the positive or negative covariance value to determine the positive or negative correlation between wind power and photovoltaic output. A positive value indicates a positive correlation, and a negative value indicates a negative correlation.

[0029] According to the analysis results of the covariance matrix, the correlation strength between wind power and photovoltaic output is output, as well as the positive or negative correlation between wind power and photovoltaic output.

[0030] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S103, includes:

[0031] Collect wind power and photovoltaic output data, wind speed, irradiance and meteorological data from wind farms and photovoltaic power stations, and extract historical output data features, wind speed features and irradiance features;

[0032] The new energy output prediction error model is used to predict the new energy output data and obtain the prediction error. The time correlation of wind power and photovoltaic power is obtained by recursive estimation of the covariance matrix and other methods. Cluster analysis is performed based on the time correlation and prediction error of the new energy output.

[0033] Assign a label to each clustering result. The label is set according to the clustering characteristics, including output stability and prediction error range.

[0034] The clustering results are sorted according to the clustering characteristics, and a corresponding relationship is established between each clustering label and the new energy output prediction model, which is used to call the corresponding prediction model during the verification process;

[0035] Use the new energy output prediction model corresponding to each cluster label to predict the data set corresponding to each cluster label, compare the prediction result with the actual output data, and obtain the comparison result between the prediction result and the actual output data;

[0036] According to the preset threshold, the valid data and invalid data in the comparison results between the prediction results and the actual output data are screened out. The prediction error less than the preset threshold is marked as valid data, and the prediction error greater than the preset threshold is marked as invalid data.

[0037] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S103, includes:

[0038] Classify and sort out the valid and invalid data obtained during the screening process, identify the invalid data, and obtain the factors that lead to inaccurate predictions, including extreme wind speeds, low irradiance, and unit failures;

[0039] Evaluate the degree of influence of factors that lead to inaccurate forecasts on forecast accuracy, determine the factors that affect the forecast results, and obtain the results of the influencing factor analysis;

[0040] According to the screening results and the results of the influencing factor analysis, the new energy output prediction model is matched, and the feature input of the model is adjusted according to the factors that lead to inaccurate predictions. The feature input of the model is adjusted according to the factors that lead to inaccurate predictions, including adding features of weather conditions and equipment status;

[0041] The new energy output data is used as a training set to train the new energy output prediction model to obtain an updated new energy output prediction model.

[0042] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S104, includes:

[0043] The reasons for the inaccurate prediction of the new energy output prediction model with prediction error are analyzed to obtain the analysis results of the reasons for the inaccurate prediction of the new energy output, which include poor data quality, mismatch of model structure and wrong parameter setting;

[0044] A deep learning model is selected to optimize the renewable energy output prediction model with prediction errors. Grid search, random search and Bayesian optimization methods are used to tune the prediction model parameters with prediction errors in turn to obtain the optimal parameter combination.

[0045] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S104, includes:

[0046] Integrate the optimized new energy output prediction model into the server-side energy management system for real-time prediction tasks;

[0047] Monitor the actual operation of the optimized new energy output prediction model, collect performance data of the new energy output prediction model in actual application, and regularly evaluate the effectiveness of the new energy output prediction model.

[0048] Furthermore, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S105, includes:

[0049] Preload the optimized new energy output prediction model, perform necessary preprocessing on the real-time data, input the preprocessed real-time data into the prediction model, and obtain the new energy output prediction result;

[0050] Analyze the forecast trend of renewable energy output in the renewable energy output forecast results, identify the peak and valley periods of output, and evaluate the uncertainty and volatility of renewable energy output by combining historical data and the current status of the real-time power system;

[0051] According to the prediction results and the current state of the power system, a matching adjustment strategy is searched in the pre-built knowledge base. The knowledge base should include load adjustment strategies, including load adjustment strategies for insufficient and excessive output of new energy in different scenarios.

[0052] Generate load adjustment instructions based on the matched adjustment strategy, which include adjusting the power consumption time of the equipment, enabling the backup power supply, and adjusting the output of the power generation equipment;

[0053] The generated adjustment instructions are sent to the corresponding power equipment through the power dispatching system, and the status of the power equipment including new energy output and load changes are monitored in real time. The feedback data of the power equipment after the execution of the adjustment instructions is collected to evaluate the effect after the execution of the adjustment instructions.

[0054] Beneficial effects of the present invention: The present invention collects and processes new energy output data, combines multiple machine learning models and covariance matrix models for accurate prediction, and significantly improves the prediction accuracy of new energy output.

[0055] Based on the prediction results of renewable energy output, the present invention can match the corresponding load adjustment strategy in real time, and automatically adjust the load output through the power dispatching system to improve the supply and demand balance and operation stability of the power system.

[0056] The present invention can adapt to the continuous changes in power system parameters and the volatility of new energy output by continuously optimizing the new energy output prediction model and load adjustment strategy, thereby improving the stability and reliability of the entire power system.

[0057] The present invention can maximize the use of new energy, reduce dependence on traditional energy, and help promote sustainable development of energy. Through automated load adjustment, the present invention reduces the need for manual intervention, reduces operating costs, and improves the operating efficiency of the power system.

[0058] The present invention integrates a large amount of historical data and real-time data, provides decision support for dispatchers through intelligent algorithms, and improves the scientificity and accuracy of decision-making. The present invention is not only applicable to large-scale wind farms and photovoltaic power stations, but can also be flexibly applied to new energy power generation facilities of various sizes, and has wide applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0060] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0062] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0063] See also Figure 1 The present invention provides an algorithm for automatically adjusting load output based on renewable energy output, including:

[0064] Step S101, obtaining new energy output data and new energy output automatic adjustment demand data, the new energy output data includes wind power output data and photovoltaic output data, the wind power output data includes wind power real-time output data, wind power historical output data, wind power predicted output data, wind power environmental data, wind turbine status data and wind farm basic information, photovoltaic output data includes photovoltaic real-time output data, photovoltaic historical output data, photovoltaic predicted output data, photovoltaic panel performance data, photovoltaic environmental data, photovoltaic power station basic information, and establishing a new energy output prediction model based on the new energy output data;

[0065] New energy output data covers two major areas: wind power and photovoltaics. Each type of data is further subdivided into multiple dimensions to ensure that the model input is as comprehensive and accurate as possible.

[0066] The present invention provides wind power output data in an algorithm for load output automatically adjusted based on new energy output, including:

[0067] Real-time wind power output data: reflects the actual power generation capacity of the wind farm at the current moment.

[0068] Wind power historical output data: Provides power generation records of wind farms over a period of time, which helps to identify output patterns and trends.

[0069] Wind power forecast output data: Based on the forecast results of meteorological, environmental and other factors, it provides the model with the expected output of future wind energy.

[0070] Wind power environment data: including wind speed, wind direction, temperature, humidity, etc., which have a direct impact on wind power output.

[0071] Wind turbine status data: reflects the operating status of the wind turbine, such as whether it is operating normally, whether there is a fault, etc.

[0072] Basic information of wind farms: such as geographical location, installed capacity, unit model, etc., helps to understand the overall operating status of the wind farm.

[0073] The present invention provides photovoltaic output data in an algorithm for automatically adjusting load output based on new energy output, including:

[0074] Photovoltaic real-time output data: reflects the actual power generation capacity of the photovoltaic power station at the current moment.

[0075] Photovoltaic historical output data: records the power generation of photovoltaic power stations over a period of time in the past, which helps to analyze output fluctuations.

[0076] Photovoltaic predicted output data: The future power generation capacity of a photovoltaic power station is predicted based on environmental factors such as irradiance and temperature.

[0077] Photovoltaic panel performance data: including the conversion efficiency and aging degree of the photovoltaic panels, which directly affect the photovoltaic output.

[0078] Photovoltaic environmental data: such as solar irradiance, ambient temperature, cloud cover, etc., have a significant impact on photovoltaic output.

[0079] Basic information of PV power stations: such as geographical location, installed capacity, and type of PV panels, helps to fully understand PV power stations.

[0080] In addition to the new energy output data, the present invention also needs to clarify the specific requirements for automatic adjustment of the new energy output, which include factors such as the current load condition of the power system, the supply and demand balance condition, and the state of the energy storage equipment.

[0081] Based on the collected renewable energy output data, a prediction model needs to be established, which involves the following steps:

[0082] Data preprocessing includes missing value processing, outlier detection and data standardization to ensure the quality and consistency of input data. According to the needs and data characteristics of new energy output forecasting, appropriate features are selected as model input.

[0083] Select a suitable model from the preset model library and use historical data as the training set for model training. The preset model library includes time series analysis, regression models, integrated learning methods or deep learning models. Evaluate model performance through cross-validation and other methods, and optimize and adjust models with inaccurate predictions.

[0084] Step S102, constructing a covariance matrix model of wind power and photovoltaic output, substituting the new energy output data into the covariance matrix model of wind power and photovoltaic output, and obtaining the time correlation information of wind power and photovoltaic output;

[0085] Extract wind power and photovoltaic output time series data rows from the database. Each time series corresponds to a variable. There are two main variables here: wind power output and photovoltaic output. Make the number and time interval of observations of each variable consistent for subsequent calculations.

[0086] Initialize an empty matrix whose dimension is equal to the length of the time series multiplied by the number of variables (2 in this invention because there are two variables: wind power and photovoltaic). The matrix will be used to store the covariance values ​​between wind power and photovoltaic output at each time point.

[0087] Traverse the time series and calculate the covariance between wind power and photovoltaic output at each time point. Covariance is a statistic that measures how two variables change together. A positive covariance indicates that the two variables have the same change trend, and a negative covariance indicates that the change trends are opposite. The size of the covariance value reflects the intensity of the change.

[0088] The preprocessed new energy output data is substituted into the covariance matrix model. The preprocessed new energy output data includes real-time data, historical data or predicted data of wind power and photovoltaic power. The wind power and photovoltaic output values ​​at each time point are used to calculate the covariance at that time point and filled into the corresponding position of the covariance matrix.

[0089] By analyzing the elements in the covariance matrix, we can identify the positive correlation, negative correlation or irrelevance between wind power and photovoltaic output. The larger the absolute value of the covariance value, the stronger the correlation between the two. A positive value indicates a positive correlation, and a negative value indicates a negative correlation.

[0090] According to the analysis results of the covariance matrix, the strength of the correlation between wind power and photovoltaic output and whether it is positive or negative correlation is crucial for subsequent clustering analysis, model optimization and formulation of load output adjustment strategies.

[0091] Step S103, clustering analysis is performed on the prediction data obtained by the new energy output prediction error model and the time correlation of wind power and photovoltaic output, to obtain the clustering result of the new energy output data, labeling and sorting the clustering result of the new energy output data, and verifying the data corresponding to each label in turn. During the verification process, a corresponding relationship is established between each label and the new energy output prediction model, and valid data and invalid data are screened out;

[0092] Collect wind power and photovoltaic output data, including historical data, real-time data and predicted data obtained through a new energy output prediction error model, and combine the wind power and photovoltaic output time correlation information obtained in step S102.

[0093] Clustering algorithms (such as K-means, hierarchical clustering, etc.) are used to perform cluster analysis on renewable energy output data. In cluster analysis, not only the numerical characteristics of renewable energy output are considered, but also the time correlation of wind power and photovoltaic power and the prediction error are combined, so that data points with similar output characteristics and prediction errors can be classified into the same category. A unique label is assigned to each clustering result, and the label setting is based on clustering characteristics, such as output stability, prediction error range, etc. Then, the clustering results are sorted according to these characteristics for subsequent verification by priority.

[0094] During the verification process, a correspondence is established between each cluster label and the new energy output prediction model. For each cluster label, one or a group of specific prediction models are used to verify the data set corresponding to the label.

[0095] The data set corresponding to each label is verified in turn, the corresponding prediction model is used to predict the data set, and the prediction result is compared with the actual output data. During the comparison, the preset threshold is used to determine whether the prediction error is within an acceptable range.

[0096] According to the comparison between the prediction results and the actual output data, the valid data and invalid data are screened out. The data with a prediction error less than the preset threshold is regarded as valid data, and vice versa. The ratio of valid data to invalid data in the data set corresponding to each cluster label and the performance of the prediction model are recorded. The characteristics and prediction errors of the invalid data are fed back to the model optimization step S104 to improve the prediction model in a targeted manner.

[0097] Step S104, optimizing the new energy output prediction model corresponding to the invalid data to obtain an optimized new energy output prediction model;

[0098] According to the results of cluster analysis and verification in step S103, the new energy output prediction models that lead to invalid prediction data are identified. These models may not be able to accurately predict the new energy output under certain conditions (such as extreme weather, equipment failure, etc.).

[0099] Cross-validation technology is used to conduct in-depth analysis on the renewable energy output prediction model corresponding to invalid data to determine the specific causes of the prediction errors. The specific causes of the prediction errors include poor data quality, mismatch of model structure, improper parameter settings, etc.

[0100] According to the reasons for the prediction errors obtained through analysis, appropriate optimization strategies are selected to improve the model. The optimization methods include the following:

[0101] Add feature inputs to the model based on the specific factors that cause inaccurate forecasts (such as extreme wind speeds, low irradiance, unit failures, etc.). For example, if the forecast error is mainly caused by extreme weather conditions, features for specific weather conditions can be added to the model.

[0102] Adjust the model structure. If the model structure mismatch is the cause of the prediction error, consider changing the model type that is more suitable for the current data. For example, switch from a simple linear regression model to a complex deep learning model.

[0103] Tune model parameters. Use grid search, random search, or Bayesian optimization to tune model parameters. Systematically explore the parameter space to find the optimal parameter combination, thereby improving the model's prediction accuracy.

[0104] Ensemble learning methods use ensemble learning techniques such as random forests and gradient boosting machines to build a strong prediction model by combining multiple weak prediction models. Ensemble learning methods can usually significantly improve the generalization ability and robustness of the model.

[0105] Use the optimized new energy output prediction model to re-predict historical data to verify the optimization effect, compare the prediction results with the actual output data, and evaluate whether the prediction accuracy of the model has been improved. If the optimization effect is significant, the optimized model will be deployed in the actual production environment.

[0106] Integrate the optimized renewable energy output prediction model into the server-side energy management system to receive new renewable energy output data in real time and make predictions, monitor the performance of the model in actual operation, and regularly evaluate its effectiveness to ensure the continued accuracy of the model.

[0107] Step S105, obtaining real-time new energy output data, substituting the real-time new energy output data into the optimized new energy output prediction model, matching in the knowledge base according to the result of the optimization algorithm, and obtaining the load output adjustment strategy.

[0108] The monitoring system of new energy power generation facilities (such as wind farms and photovoltaic power stations) can be used to obtain wind power and photovoltaic output data in real time, including real-time wind power output, photovoltaic output and related environmental parameters (such as wind speed, irradiance, etc.).

[0109] Perform necessary preprocessing on the acquired real-time data, including missing value processing, outlier detection, and data standardization, to ensure the quality and consistency of the input data.

[0110] The preprocessed real-time renewable energy output data is substituted into the optimized renewable energy output prediction model, and the model is used for real-time prediction to obtain the prediction results of renewable energy output in the future.

[0111] Based on the forecast results, the forecast trend of renewable energy output is analyzed, the peak and trough periods of output are identified, and the uncertainty and volatility of renewable energy output are evaluated by combining historical data and the current status of the real-time power system.

[0112] According to the prediction results and the current state of the power system, a matching search is performed in the pre-built knowledge base to obtain a suitable load output adjustment strategy. The knowledge base should include various load adjustment strategies to cope with insufficient or excessive output of new energy in different scenarios.

[0113] According to the matched adjustment strategy, specific load adjustment instructions are generated. The load adjustment instructions include adjusting the power consumption time of the equipment, enabling the backup power supply, adjusting the output of the power generation equipment, etc.

[0114] The generated adjustment instructions are sent to relevant equipment through the power dispatching system to monitor the status of the power system (including new energy output, load changes, etc.) in real time to ensure the effective implementation of the adjustment strategy.

[0115] During the implementation of the adjustment strategy, the state changes of the power system are continuously monitored, including the output of new energy, load demand, and equipment operating status. The feedback data of the adjusted system is collected, including the actual output of new energy, load changes, and equipment operating status. Based on the collected feedback data, the implementation effect of the adjustment strategy is evaluated. If the effect is not ideal, the subsequent strategy is adjusted and optimized as needed. Based on the evaluation results and new real-time data, the new energy output prediction model and load adjustment strategy are continuously optimized to adapt to the continuous changes in power system parameters and the volatility of new energy output.

[0116] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S101, includes:

[0117] Preprocessing the new energy output data to obtain preprocessed new energy output data;

[0118] receiving the influencing characteristics of the new energy output forecast and the new energy output forecast demand, the influencing characteristics of the new energy output forecast including time, weather conditions and equipment status;

[0119] Matching machine learning models in a preset model library according to the forecast demand for new energy output. The machine learning models include time series analysis, regression models, integrated learning methods or deep learning models;

[0120] Using historical data as a training set, the influencing features of renewable energy output prediction are input into the matched machine learning model, and model training is performed to obtain a renewable energy output prediction model.

[0121] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S101, includes:

[0122] Deploy the trained new energy output prediction model to the server, receive new new energy output data as input data in real time, and output the prediction results of new energy output;

[0123] Using the newly acquired new energy output data according to a preset time period to construct a new energy output prediction model to obtain an updated new energy output prediction model;

[0124] A simulation mirror data processing area is established, and the new energy output data is copied to the simulation mirror data processing area. The new energy output data in the simulation mirror data processing area is predicted by the updated new energy output prediction model to obtain the prediction result of the updated new energy output prediction model. The prediction result of the updated new energy output prediction model is compared with the actual output data in the new energy output data. If the actual output data in the new energy output data is consistent with the prediction result of the updated new energy output prediction model, the updated new energy output prediction model replaces the existing new energy output prediction model.

[0125] The trained new energy output prediction model is deployed on a high-performance server to enable it to quickly respond to and process large amounts of real-time data. The server receives real-time output data from different new energy sites (such as wind farms and photovoltaic power stations), including the real-time output of wind power and photovoltaic power, as well as related environmental parameters (such as wind speed, irradiance, etc.).

[0126] The new energy output prediction model receives new data input, performs prediction calculations, and outputs prediction results of the new energy output.

[0127] In order to maintain the prediction accuracy of the model, the model update process is automatically started according to a preset time period (such as daily, weekly or monthly).

[0128] In each update cycle, the latest acquired new energy output data is collected, which integrates the latest real-time data, historical records and possible forecast data.

[0129] The newly collected new energy output data is used to reconstruct the new energy output prediction model, and the updated prediction model is obtained by repeating the training process in step S101.

[0130] In order to evaluate the performance of the updated renewable energy output prediction model, a simulated mirror data processing area was created that replicated the actual data processing environment.

[0131] New energy output data (including historical and real-time data) is copied to this simulation area to ensure the authenticity and consistency of the test data.

[0132] The updated new energy output prediction model predicts the data in this simulation environment and generates prediction results.

[0133] Subsequently, the prediction results will be compared with the actual new energy output data, and the accuracy of the model will be evaluated by calculating the prediction error. If the updated model performs well in the test in the simulation environment, that is, its prediction results are highly consistent with the actual output data and meet the accuracy standards preset by the system, then the system will make a decision to replace the currently used model. On the premise of ensuring the stable operation of the entire system, the new energy output prediction model currently in use will be replaced by the updated model. During the replacement process, the model version information before and after the replacement is recorded in detail to facilitate the subsequent tracking of model performance and tracing of problems.

[0134] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S102, includes:

[0135] Extract wind power and photovoltaic output time series data rows from the database, determine the time series of wind power and photovoltaic output, each time series corresponds to a variable, the variables include wind power output and photovoltaic output, determine the number of observations and time intervals of each variable, so that the length of each variable time series and the sampling rate of each variable are consistent;

[0136] Initialize an empty matrix, the dimension of the initialized matrix is ​​equal to the length of the time series multiplied by the number of variables;

[0137] Calculate the covariance between wind power and photovoltaic output at the time point corresponding to each element in the matrix to obtain the covariance calculation result between wind power and photovoltaic output;

[0138] Analyze the elements in the covariance calculation results between wind power and photovoltaic output, identify the positive correlation, negative correlation or irrelevance between wind power and photovoltaic output, and judge the strength of the correlation according to the size of the elements in the covariance calculation results between wind power and photovoltaic output.

[0139] When extracting wind power and photovoltaic output time series data from the database, the integrity and consistency of the acquired data must be ensured. Any missing data points can be filled by interpolation or other appropriate statistical methods.

[0140] In the process of extracting data, it is necessary to verify whether the number of observations of wind power and photovoltaic output time series is the same and whether the time intervals are consistent. If there is inconsistency, the data needs to be resampled or interpolated so that the two time series have corresponding observations at the same time point.

[0141] When initializing an empty matrix, you need to calculate the dimension of the matrix. If the time series lengths of the two variables, wind power and photovoltaic power, are both N, then theoretically the dimension of the covariance matrix should be N × N. However, in this scenario, we are concerned about the covariance between the two variables, so we actually need a 2 × 2 matrix to store the covariance values ​​between the two variables.

[0142] The formula for calculating the covariance is: textCov(X,Y)=fracsumi=1n(Xi−barX)(Yi−barY)n−1, where X and Y represent the output time series of wind power and photovoltaic power, respectively, barX and barY are their average values, and n represents the total number of observations.

[0143] In the actual calculation process, it is necessary to traverse the entire time series and apply the above formula to calculate the wind power and photovoltaic output values ​​at each time point.

[0144] The covariance value obtained can reflect the strength and direction of the linear correlation between wind power and photovoltaic output. A positive value indicates a positive correlation, that is, when one variable increases, the other variable tends to increase; a negative value indicates a negative correlation; and a value close to 0 means that the linear correlation between the two is weak or non-existent.

[0145] The absolute value of the covariance value can also be used to determine the strength of the correlation. The larger the absolute value, the stronger the linear relationship between the two.

[0146] After completing the analysis of the covariance matrix (or a single covariance value), these results can be applied to subsequent clustering analysis, model optimization, or strategy formulation.

[0147] For example, if a strong positive correlation is found between wind power and photovoltaic output, this synchronicity can be considered when formulating load adjustment strategies, thereby more effectively achieving supply and demand balance management of the power system.

[0148] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S102, includes:

[0149] Each element in the covariance matrix represents the covariance of wind power and photovoltaic output at a certain point in time. The size of the covariance value is used to reflect the strength of the correlation between wind power and photovoltaic output. The larger the absolute value of the covariance value, the stronger the correlation between wind power and photovoltaic output.

[0150] Determine the positive or negative covariance value to determine the positive or negative correlation between wind power and photovoltaic output. A positive value indicates a positive correlation, and a negative value indicates a negative correlation.

[0151] According to the analysis results of the covariance matrix, the correlation strength between wind power and photovoltaic output is output, as well as the positive or negative correlation between wind power and photovoltaic output.

[0152] Before covariance analysis, it is necessary to accurately extract the wind power and photovoltaic output time series data from the database, while ensuring the integrity and accuracy of the data. The extracted data needs to be preprocessed, including processing missing data points, detecting and processing outliers, and performing data standardization.

[0153] According to the length and sampling rate of wind power and photovoltaic time series, a two-dimensional covariance matrix is ​​initialized. The dimension of this matrix is ​​equal to the length of the time series multiplied by 2, because the data of two variables, wind power and photovoltaic, are to be analyzed.

[0154] Each element in the matrix is ​​initially set to 0 and is used to store the calculated covariance values.

[0155] Traversing the time series of wind power and photovoltaic power, for each pair of wind power output and photovoltaic output data at corresponding time points, the covariance formula is used to calculate the correlation between them.

[0156] The covariance calculation formula is used to quantify the linear relationship between two variables (such as wind power output and photovoltaic output). The covariance calculation formula is as follows:

[0157] Cov(X,Y) = \frac{\sum_{i=1}^{n}(Xi-\bar{X})(Yi-\bar{Y})}{n-1}

[0158] in:

[0159] Cov(X,Y) is the covariance between variables X and Y.

[0160] Xi and Yi are the ith observations of variables X and Y respectively.

[0161] \bar{X} and \bar{Y} are the means of variables X and Y respectively.

[0162] n is the total number of observations.

[0163] This formula is used to calculate the covariance between wind power output and photovoltaic output at each time point. The size of the covariance value can reflect the strength and direction of the linear correlation between the two. The positive or negative covariance value indicates the positive or negative correlation between the variables, and the absolute value of the covariance value reflects the strength of the correlation.

[0164] The calculated covariance values ​​will be filled into the corresponding positions of the previously initialized covariance matrix. It is worth noting that the elements on the diagonal of the matrix actually represent the variance of the wind power or photovoltaic output itself.

[0165] By analyzing the elements in the covariance matrix, the strength and direction of the correlation between wind power and photovoltaic output can be determined.

[0166] The output results include the strength of the correlation between wind power and PV output (measured by the absolute value of the covariance value) and whether they are positively or negatively correlated (determined by the sign of the covariance value).

[0167] The correlation information obtained in step S102 will be used in subsequent steps, such as cluster analysis or model optimization, to improve the accuracy of new energy output prediction.

[0168] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S103, includes:

[0169] Collect wind power and photovoltaic output data, wind speed, irradiance and meteorological data from wind farms and photovoltaic power stations, and extract historical output data features, wind speed features and irradiance features;

[0170] The new energy output prediction error model is used to predict the new energy output data and obtain the prediction error. The time correlation of wind power and photovoltaic power is obtained by recursive estimation of the covariance matrix and other methods. Cluster analysis is performed based on the time correlation and prediction error of the new energy output.

[0171] Assign a label to each clustering result. The label is set according to the clustering characteristics, including output stability and prediction error range.

[0172] The clustering results are sorted according to the clustering characteristics, and a corresponding relationship is established between each clustering label and the new energy output prediction model, which is used to call the corresponding prediction model during the verification process;

[0173] Use the new energy output prediction model corresponding to each cluster label to predict the data set corresponding to each cluster label, compare the prediction result with the actual output data, and obtain the comparison result between the prediction result and the actual output data;

[0174] According to the preset threshold, the valid data and invalid data in the comparison results between the prediction results and the actual output data are screened out. The prediction error less than the preset threshold is marked as valid data, and the prediction error greater than the preset threshold is marked as invalid data.

[0175] Collect wind power and photovoltaic output data from wind farms and photovoltaic power stations in real time, and record wind speed, irradiance and related meteorological data. At the same time, extract historical output data, wind speed and irradiance data from the historical database, extract the characteristics of these data, and prepare for subsequent analysis and modeling.

[0176] The pre-trained new energy output prediction error model is used to make a preliminary prediction of the new energy output data to obtain preliminary prediction data. Through methods such as recursive estimation of the covariance matrix, the time correlation between wind power and photovoltaic power at different time points is deeply analyzed, which is convenient for analyzing the synchronization or asynchrony of the output changes of the two.

[0177] Next, clustering algorithms (such as K-means or hierarchical clustering) are used to perform cluster analysis on the renewable energy output data. When clustering, the output value, time correlation, prediction error, wind speed, irradiance and other characteristics are comprehensively considered.

[0178] Each clustering result is assigned a unique label, which is set based on the clustering characteristics and reflects information such as output stability and prediction error range.

[0179] The clustering results are sorted according to clustering characteristics such as prediction accuracy and stability. A corresponding new energy output prediction model is matched to each cluster label, and the new energy output prediction model is optimized for the characteristics of this type of data during training.

[0180] During the validation process, these models are used to make predictions on the data sets under the corresponding labels and compared with the actual output data.

[0181] A preset error threshold is used to determine whether the deviation between the prediction result and the actual output data is within an acceptable range. Data with a prediction error less than this threshold is considered valid data, otherwise it is considered invalid data.

[0182] The characteristics and prediction errors of invalid data are fed back to the model optimization step for targeted improvement. By analyzing the causes of invalid data (such as extreme weather, equipment failure, or data quality issues), the model's feature inputs, structure, or parameters can be adjusted to improve prediction accuracy.

[0183] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S103, includes:

[0184] Classify and sort out the valid and invalid data obtained during the screening process, identify the invalid data, and obtain the factors that lead to inaccurate predictions, including extreme wind speeds, low irradiance, and unit failures;

[0185] Evaluate the degree of influence of factors that lead to inaccurate forecasts on forecast accuracy, determine the factors that affect the forecast results, and obtain the results of the influencing factor analysis;

[0186] According to the screening results and the results of the influencing factor analysis, the new energy output prediction model is matched, and the feature input of the model is adjusted according to the factors that lead to inaccurate predictions. The feature input of the model is adjusted according to the factors that lead to inaccurate predictions, including adding features of weather conditions and equipment status;

[0187] The new energy output data is used as a training set to train the new energy output prediction model to obtain an updated new energy output prediction model.

[0188] The key is to conduct in-depth feature analysis of invalid data, covering the time series characteristics of output data, changes in environmental parameters, and abnormalities in equipment status parameters. Comparing the feature differences between valid and invalid data can more accurately identify the specific reasons for forecast errors.

[0189] Visualization tools use tools such as Matplotlib and Seaborn to visualize invalid data features, thereby more intuitively showing the difference between invalid and valid data and helping to identify inaccurate prediction patterns.

[0190] Quantify each identified influencing factor (extreme wind speed, low irradiance, unit failure) to determine their specific contribution to the forecast error. This can be achieved through sensitivity analysis, variance decomposition and other methods. The interaction between multiple influencing factors needs to be considered. Use methods such as multivariate regression analysis and principal component analysis to evaluate the impact of multiple factors on forecast accuracy.

[0191] Select or construct new features based on the results of the influencing factor analysis. For example, for extreme wind speed, you can construct features such as wind speed extreme value index and wind speed change rate; for low irradiance, you can introduce features such as cloud thickness and sky brightness. Perform engineering steps such as feature scaling and encoding to ensure that the new features are consistent in scale with the original features so that the model can process them correctly.

[0192] K-fold cross validation is used when training the model to ensure the generalization ability of the model on different data subsets. In addition to common indicators such as MSE and RMSE, R², MAE and other multi-angle evaluations of model performance can also be used. The trained model needs to be saved and deployed to the server so that new data can be received in real time for prediction. After deployment, the model performance needs to be monitored in real time to ensure its stability and accuracy. A feedback mechanism is established to collect prediction results and actual output data in actual applications, and to regularly evaluate and optimize the model.

[0193] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S104, includes:

[0194] The reasons for the inaccurate prediction of the new energy output prediction model with prediction error are analyzed to obtain the analysis results of the reasons for the inaccurate prediction of the new energy output, which include poor data quality, mismatch of model structure and wrong parameter setting;

[0195] A deep learning model is selected to optimize the renewable energy output prediction model with prediction errors. Grid search, random search and Bayesian optimization methods are used to tune the prediction model parameters with prediction errors in turn to obtain the optimal parameter combination.

[0196] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S104, includes:

[0197] Integrate the optimized new energy output prediction model into the server-side energy management system for real-time prediction tasks;

[0198] Monitor the actual operation of the optimized new energy output prediction model, collect performance data of the new energy output prediction model in actual application, and regularly evaluate the effectiveness of the new energy output prediction model.

[0199] Automatically identify and filter out new energy output forecast models that have forecast errors, and conduct detailed analysis of the causes of inaccurate forecasts, including:

[0200] Check for missing, anomalous, or noisy data, which can affect forecast accuracy.

[0201] Evaluate whether the model is suitable for processing specific renewable energy output data and whether the model complexity matches the data characteristics.

[0202] Analyze whether the parameter settings are reasonable and whether there is overfitting or underfitting.

[0203] Select a suitable deep learning model for optimization, such as RNN, LSTM, or CNN, and use a variety of methods to tune parameters:

[0204] Grid Search: Systematically iterates over parameter combinations, identifying the best settings via cross-validation.

[0205] Random search: randomly sample in parameter space to explore more combinations.

[0206] Bayesian Optimization: Use Bayesian statistics to build a proxy model to find a near-global optimal solution with fewer iterations.

[0207] Integrate the optimized model into the server-side energy management system to achieve real-time prediction, monitoring and evaluation:

[0208] Real-time monitoring and evaluation: Continuously monitor the actual performance of the model, collect comparison information between prediction and actual data, and evaluate the real-time prediction effect.

[0209] Feedback and optimization loop: If performance degrades, the system will trigger a new round of optimization, including data inspection, model adjustment, parameter tuning, etc., to ensure high prediction accuracy.

[0210] Specifically, the algorithm for automatically adjusting the load output based on the output of new energy provided by the present invention, step S105, includes:

[0211] Preload the optimized new energy output prediction model, perform necessary preprocessing on the real-time data, input the preprocessed real-time data into the prediction model, and obtain the new energy output prediction result;

[0212] Analyze the forecast trend of renewable energy output in the renewable energy output forecast results, identify the peak and valley periods of output, and evaluate the uncertainty and volatility of renewable energy output by combining historical data and the current status of the real-time power system;

[0213] According to the prediction results and the current state of the power system, a matching adjustment strategy is searched in the pre-built knowledge base. The knowledge base should include load adjustment strategies, including load adjustment strategies for insufficient and excessive output of new energy in different scenarios.

[0214] Generate load adjustment instructions based on the matched adjustment strategy, which include adjusting the power consumption time of the equipment, enabling the backup power supply, and adjusting the output of the power generation equipment;

[0215] The generated adjustment instructions are sent to the corresponding power equipment through the power dispatching system, and the status of the power equipment including new energy output and load changes are monitored in real time. The feedback data of the power equipment after the execution of the adjustment instructions is collected to evaluate the effect after the execution of the adjustment instructions.

[0216] After receiving real-time data, data cleaning is first required to remove or correct outliers and missing values ​​to ensure the quality of data input into the prediction model. In order to maintain the stability of the model input, the real-time data needs to be standardized or normalized according to the statistical characteristics of historical data.

[0217] Extract key time series features such as trends, seasonality, and periodicity from real-time data to improve the accuracy of the forecast model. Design an efficient data input interface to receive real-time monitoring data of renewable energy output. Determine the batch size and processing frequency of real-time forecasts based on system requirements to balance the timeliness and accuracy of forecasts. Cache the forecast results and output them to the load adjustment system in real time through the preset interface.

[0218] Use statistical methods or machine learning algorithms to analyze the changing trends of renewable energy output. Use the concept of prediction interval to quantify the uncertainty and volatility of renewable energy output. Compare real-time prediction results with historical data to evaluate the accuracy of the prediction.

[0219] Build a structured knowledge base containing various load adjustment strategies and their applicable conditions. Design an efficient strategy matching algorithm to quickly find the most appropriate adjustment strategy based on the prediction results and power system status. Among multiple matching strategies, sort them according to effectiveness and feasibility to prioritize the optimal strategy.

[0220] The selected adjustment strategy is formatted into a specific load adjustment instruction, and the adjustment instruction is sent to the corresponding power equipment safely and reliably through the power dispatching system. The instruction execution process is monitored in real time and the equipment feedback data is collected to evaluate the adjustment effect. Reasonable evaluation indicators are designed to quantify the effect of the adjustment strategy, such as supply and demand balance, new energy utilization rate, etc.

[0221] The technical solution of the present invention solves the problem that the automatic adjustment of renewable energy output is difficult to accurately predict and control due to the dynamic changes of multiple parameters such as voltage, current, power, and the mutual influence between the parameters in the operation of the power system. The specific solution is as follows:

[0222] The present invention firstly collects comprehensive new energy output data, including real-time, historical and predicted output data of wind power and photovoltaic power, as well as relevant environmental data and equipment status data.

[0223] The collected renewable energy output data are preprocessed, including missing value processing, outlier detection and data standardization, to ensure the quality and consistency of the input data.

[0224] According to the needs of new energy output forecasting, suitable machine learning models are matched in the preset model library. The machine learning models include time series analysis, regression models, integrated learning methods or deep learning models.

[0225] The selected model is trained using historical data as a training set to obtain a new energy output prediction model. The prediction accuracy is improved by continuously iterating and updating the model.

[0226] A covariance matrix model of wind power and photovoltaic output is constructed to analyze the time correlation between wind power and photovoltaic output. Through the covariance matrix, it can be clarified whether the output change trends of wind power and photovoltaic at different time points are consistent or opposite, so as to more accurately predict the output of new energy.

[0227] The time correlation of wind power and photovoltaic power is clustered with the forecast data obtained by the new energy output forecast error model to identify data groups with different output stability and forecast error range.

[0228] The clustering results are verified to screen out valid and invalid data, and the new energy output prediction model corresponding to the invalid data is optimized. The prediction accuracy of the model is improved by adding specific weather conditions, equipment status and other features.

[0229] Deploy the optimized renewable energy output prediction model to the server side, receive new renewable energy output data in real time and make predictions. According to the real-time prediction results and the current status of the power system, find the matching load adjustment strategy in the knowledge base. The load adjustment strategy includes the load adjustment strategy for insufficient and excessive renewable energy output in different scenarios. Generate load adjustment instructions and send them to related equipment through the power dispatching system, monitor the power system status in real time, and ensure the effective execution of the adjustment strategy.

[0230] Regularly evaluate the accuracy of the forecast model and the effectiveness of the load adjustment strategy, and collect system feedback data. Based on the evaluation results and new data, continuously optimize the renewable energy output forecast model and load adjustment strategy to adapt to the continuous changes in power system parameters and the volatility of renewable energy output.

[0231] In summary, the present invention effectively solves the problem of accurate prediction and control in the automatic adjustment of renewable energy output in the power system by comprehensively collecting and processing renewable energy output data, combining multiple machine learning models and covariance matrix models for accurate prediction, as well as real-time strategy adjustment and continuous optimization.

[0232] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations, and the above-described embodiments of the present invention do not constitute a limitation on the scope of protection of the present invention.

Claims

1. An algorithm for automatically adjusting load output based on renewable energy output, characterized in that: include: Step S101, obtaining new energy output data and new energy output automatic adjustment demand data, the new energy output data includes wind power output data and photovoltaic output data, the wind power output data includes wind power real-time output data, wind power historical output data, wind power predicted output data, wind power environmental data, wind turbine status data and wind farm basic information, photovoltaic output data includes photovoltaic real-time output data, photovoltaic historical output data, photovoltaic predicted output data, photovoltaic panel performance data, photovoltaic environmental data, photovoltaic power station basic information, and establishing a new energy output prediction model based on the new energy output data; Step S102, constructing a covariance matrix model of wind power and photovoltaic output, substituting the new energy output data into the covariance matrix model of wind power and photovoltaic output, and obtaining the time correlation information of wind power and photovoltaic output; Step S103, clustering analysis is performed on the prediction data obtained by the new energy output prediction error model and the time correlation of wind power and photovoltaic output, to obtain the clustering result of the new energy output data, labeling and sorting the clustering result of the new energy output data, and verifying the data corresponding to each label in turn. During the verification process, a corresponding relationship is established between each label and the new energy output prediction model, and valid data and invalid data are screened out; Step S104, optimizing the new energy output prediction model corresponding to the invalid data to obtain an optimized new energy output prediction model; Step S105, obtaining real-time new energy output data, substituting the real-time new energy output data into the optimized new energy output prediction model, matching in the knowledge base according to the result of the optimization algorithm, and obtaining the load output adjustment strategy; Step S101 includes: Preprocessing the new energy output data to obtain preprocessed new energy output data; receiving the influencing characteristics of the new energy output forecast and the new energy output forecast demand, wherein the influencing characteristics of the new energy output forecast include time, weather conditions, and equipment status; Matching machine learning models in a preset model library according to the forecast demand for new energy output. The machine learning models include time series analysis, regression models, integrated learning methods or deep learning models; Use historical data as a training set, input the influencing features of renewable energy output prediction into the matched machine learning model, and perform model training to obtain a renewable energy output prediction model; Step S103 includes: Collect wind power and photovoltaic output data, wind speed, irradiance and meteorological data from wind farms and photovoltaic power stations, and extract historical output data features, wind speed features and irradiance features; The new energy output prediction error model is used to predict the new energy output data and obtain the prediction error. The time correlation of wind power and photovoltaic power is obtained by recursive estimation of the covariance matrix method. Cluster analysis is performed based on the time correlation and prediction error of the new energy output. Assign a label to each clustering result. The label is set according to the clustering characteristics, including output stability and prediction error range. The clustering results are sorted according to the clustering characteristics, and a corresponding relationship is established between each clustering label and the new energy output prediction model, which is used to call the corresponding prediction model during the verification process; Use the new energy output prediction model corresponding to each cluster label to predict the data set corresponding to each cluster label, compare the prediction result with the actual output data, and obtain the comparison result between the prediction result and the actual output data; According to the preset threshold, the valid data and invalid data in the comparison result between the prediction result and the actual output data are screened out. The prediction error less than the preset threshold is marked as valid data, and the prediction error greater than the preset threshold is marked as invalid data; Classify and sort out the valid and invalid data obtained during the screening process, identify the invalid data, and obtain the factors that lead to inaccurate predictions, including extreme wind speeds, low irradiance, and unit failures; Evaluate the degree of influence of factors that lead to inaccurate predictions on the accuracy of predictions, determine the factors that affect the prediction results, and obtain the results of the influencing factor analysis; According to the screening results and the results of the influencing factor analysis, the new energy output prediction model is matched, and the feature input of the model is adjusted according to the factors that lead to inaccurate predictions. The feature input of the model is adjusted according to the factors that lead to inaccurate predictions, including adding features of weather conditions and equipment status; Using the new energy output data as a training set, the new energy output prediction model is trained to obtain an updated new energy output prediction model; Step S104 includes: The reasons for the inaccurate prediction of the new energy output prediction model with prediction error are analyzed to obtain the analysis results of the reasons for the inaccurate prediction of the new energy output, which include poor data quality, mismatch of model structure and wrong parameter setting; A deep learning model is selected to optimize the new energy output prediction model with prediction errors, and grid search, random search and Bayesian optimization methods are used to tune the prediction model parameters with prediction errors in turn to obtain the optimal parameter combination; Integrate the optimized new energy output prediction model into the server-side energy management system for real-time prediction tasks; Monitor the actual operation of the optimized new energy output prediction model, collect performance data of the new energy output prediction model in actual application, and regularly evaluate the effectiveness of the new energy output prediction model.

2. The algorithm for automatically adjusting the load output based on the output of new energy sources as claimed in claim 1 is characterized in that: Step S101 includes: Deploy the trained new energy output prediction model to the server, receive new new energy output data as input data in real time, and output the prediction results of new energy output; Using the newly acquired new energy output data according to a preset time period to construct a new energy output prediction model to obtain an updated new energy output prediction model; A simulation mirror data processing area is established, and the new energy output data is copied to the simulation mirror data processing area. The new energy output data in the simulation mirror data processing area is predicted by the updated new energy output prediction model to obtain the prediction result of the updated new energy output prediction model. The prediction result of the updated new energy output prediction model is compared with the actual output data in the new energy output data. If the actual output data in the new energy output data is consistent with the prediction result of the updated new energy output prediction model, the updated new energy output prediction model replaces the existing new energy output prediction model.

3. The algorithm for automatically adjusting the load output based on the output of new energy sources as claimed in claim 1, characterized in that: Step S102 includes: Extract wind power and photovoltaic output time series data rows from the database, determine the time series of wind power and photovoltaic output, each time series corresponds to a variable, the variables include wind power output and photovoltaic output, determine the number of observations and time intervals of each variable, so that the length of each variable time series and the sampling rate of each variable are consistent; Initialize an empty matrix, the dimension of the initialized matrix is ​​equal to the length of the time series multiplied by the number of variables; Calculate the covariance between wind power and photovoltaic output at the time point corresponding to each element in the matrix to obtain the covariance calculation result between wind power and photovoltaic output; Analyze the elements in the covariance calculation results between wind power and photovoltaic output, identify the positive correlation, negative correlation or irrelevance between wind power and photovoltaic output, and judge the strength of the correlation according to the size of the elements in the covariance calculation results between wind power and photovoltaic output.

4. The algorithm for automatically adjusting the load output based on the output of new energy sources as claimed in claim 3 is characterized in that: Step S102 includes: Each element in the covariance matrix represents the covariance of wind power and photovoltaic output at a certain point in time. The size of the covariance value is used to reflect the strength of the correlation between wind power and photovoltaic output. The larger the absolute value of the covariance value, the stronger the correlation between wind power and photovoltaic output. Determine the positive or negative covariance value to determine the positive or negative correlation between wind power and photovoltaic output. A positive value indicates a positive correlation, and a negative value indicates a negative correlation. According to the analysis results of the covariance matrix, the correlation strength between wind power and photovoltaic output is output, as well as the positive or negative correlation between wind power and photovoltaic output.

5. The algorithm for automatically adjusting load output based on renewable energy output as claimed in claim 1, characterized in that: Step S105 includes: Preload the optimized new energy output prediction model, perform necessary preprocessing on the real-time data, input the preprocessed real-time data into the prediction model, and obtain the new energy output prediction result; Analyze the forecast trend of renewable energy output in the renewable energy output forecast results, identify the peak and valley periods of output, and evaluate the uncertainty and volatility of renewable energy output by combining historical data and the current status of the real-time power system; According to the prediction results and the current state of the power system, a matching adjustment strategy is searched in the pre-built knowledge base. The knowledge base should include load adjustment strategies, including load adjustment strategies for insufficient and excessive output of new energy in different scenarios. Generate load adjustment instructions based on the matched adjustment strategy, which include adjusting the power consumption time of the equipment, enabling the backup power supply, and adjusting the output of the power generation equipment; The generated adjustment instructions are sent to the corresponding power equipment through the power dispatching system, and the status of the power equipment including new energy output and load changes are monitored in real time. The feedback data of the power equipment after the execution of the adjustment instructions is collected to evaluate the effect after the execution of the adjustment instructions.

Images