Power grid multi-energy data acquisition frequency control method considering load prediction error

Abstract

The invention provides a power grid multi-energy data acquisition frequency control method considering a load prediction error, and belongs to the technical field of prediction control. The method comprises the steps of determining energy types of multiple energy sources of a research power grid, collecting original data of the energy sources according to energy characteristics of any energy type, and standardizing the original data to obtain standard power generation data; building a load prediction model, taking the standard power generation data of all energy types as the input of the load prediction model, and outputting a load prediction result; performing error analysis on the load prediction result and the actual load result to obtain an error coefficient of the corresponding energy type; based on a control algorithm, combining the error coefficient of the energy type and standard power generation data, and adjusting the collection frequency of the energy; and evaluating the frequency control result, and reversely optimizing the control algorithm and the load prediction model based on the evaluation result. And the load prediction precision and scheduling efficiency of the multi-energy power grid are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of predictive control, and particularly to a control method for the data acquisition frequency of multi-energy in a power grid considering load prediction errors. Background Art

[0002] At present, with the transformation of the global energy structure and the opening of the power market, more and more renewable energy sources (such as wind energy and solar energy) are incorporated into the power system. However, renewable energy sources such as wind energy and solar energy are volatile and uncertain, which poses higher challenges to the stable operation of the power grid. In particular, the multi-energy hybrid scheduling in the power grid and the accuracy of load prediction have become one of the key factors affecting the operation efficiency and stability of the power grid. However, the data acquisition frequencies of different energy types are different, resulting in a mismatch between the timeliness and accuracy of the input data of the load prediction model.

[0003] Therefore, the present invention proposes a control method for the data acquisition frequency of multi-energy in a power grid considering load prediction errors. Summary of the Invention

[0004] The present invention provides a control method for the data acquisition frequency of multi-energy in a power grid considering load prediction errors, which is used to improve the accuracy of power grid scheduling by gradually optimizing the data acquisition and load prediction processes of the multi-energy system in the power grid. First, determine different energy types and standardize the original power generation data; then, build a load prediction model, analyze the errors and obtain the error coefficients; then, adjust the acquisition frequency through a control algorithm to optimize the data acquisition process; finally, reverse-optimize the control algorithm and the prediction model based on the evaluation results to ensure the efficiency and stability of the system. This method can effectively improve the accuracy and operability of multi-energy scheduling in the power grid.

[0005] On the one hand, the present invention provides a control method for the data acquisition frequency of multi-energy in a power grid considering load prediction errors, including:

[0006] Step 1: Determine the energy types of the multi-energy in the power grid under study, and collect the original data of the energy according to the energy characteristics of any energy type, and standardize the original data to obtain standard power generation data;

[0007] Step 2: Build a load prediction model, use the standard power generation data of all energy types as the input of the load prediction model, and output the load prediction result;

[0008] Step 3: Conduct error analysis on the load prediction result and the actual load result to obtain the error coefficients corresponding to the energy types;

[0009] Step 4: Based on the control algorithm, combine the error coefficients of the energy types and the standard power generation data to adjust the acquisition frequency of the energy;

[0010] Step 5: Evaluate the frequency control results and reverse-optimize the control algorithm and load prediction model based on the evaluation results.

[0011] On the other hand, determine the energy types of the multi-energy of the research power grid, and collect the original data of the energy according to the energy characteristics of any energy type, including:

[0012] According to the configuration information of the target research power grid, obtain all the energy types connected to the target research power grid;

[0013] Select the corresponding sensor type according to the energy type, and identify the installation positions of all sensors according to the three-dimensional geographical map of the target research power grid;

[0014] Based on the energy characteristics of any energy type, obtain the matching scores of the corresponding sensor and any installation position, select the installation position with the highest matching score of the sensor as the optimal installation position, configure a unique first number for the optimal installation position, and configure a unique second number for the sensor;

[0015] According to the corresponding relationship between the first number and the second number, complete the installation configuration of all sensors, and collect the original data of the corresponding energy type based on the sensors.

[0016] On the other hand, standardize the original data to obtain standard power generation data, including:

[0017] Obtain the original data of all energy types, perform time alignment processing on the original data of any energy type according to the preset standard time interval to obtain the first power generation data;

[0018] Identify the missing values in the first power generation data, and fill the missing values based on zero mean and unit variance to obtain the standard power generation data;

[0019] Initialize the power generation database according to the preset field type, use the energy type and collection time as the combined primary key, and store the standard power generation data in the power generation database.

[0020] On the other hand, before building the load prediction model, use feature engineering to process the standard power generation data to obtain energy power generation characteristics, including:

[0021] Based on the power generation database, obtain the standard power generation data of all energy types, and use feature engineering to process the standard power generation data to obtain energy power generation characteristics;

[0022] According to the standard power generation data, construct the feature matrix L:

[0023] where, a ijData of the j-th time node of the i-th energy type, where n represents that each energy type has n time nodes, and m represents that there are m energy types in total;

[0024] Construct the divergence matrix B as: where, μ k represents the mean vector of the k-th energy type, represents the mean of all standard power generation data, T represents the transpose symbol, and · represents the dot product symbol;

[0025] Based on the feature matrix L and the divergence matrix B, and using the maximization function, the projected feature matrix W is obtained:

[0026] Maximize the variance of the between-class divergence matrix B in the low-dimensional space, and at the same time minimize the variance of the within-class feature matrix L in the low-dimensional space to obtain the optimal projected feature matrix W;

[0027] According to L = W·ω, the eigenvector is obtained

[0028] where, ω1 represents the power generation feature of the first energy type, ω2 represents the power generation feature of the second energy type, and ω m represents the power generation feature of the m-th energy type.

[0029] On the other hand, build a load forecasting model, and use the standard power generation data of all energy types as the input of the load forecasting model to output the load forecasting results, including:

[0030] Build the framework of the load forecasting model according to the preset forecasting model, obtain the historical power generation data by matching the data in the historical database according to the energy type, use the historical power generation data as the training set, and use the energy power generation feature as the validation set;

[0031] Based on the training set, perform iterative training on the preset forecasting model to obtain the first model parameters, use the first model parameters as the framework nodes to obtain the first load forecasting model;

[0032] Based on the validation set, verify the first load forecasting model to obtain the model hyperparameters, and optimize and adjust the first model parameters according to the Bayesian optimization algorithm combined with the hyperparameters to obtain the final model parameters;

[0033] Input the final model parameters into the first load forecasting model to obtain the final load forecasting model;

[0034] Use the standard power generation data of all energy types as the input of the load forecasting model to output the load forecasting results.

[0035] On the other hand, the error analysis is carried out on the load prediction results and the actual load results to obtain the error coefficients corresponding to the energy types, including:

[0036] The error analysis is carried out on the load prediction results and the actual load results of any energy type to obtain the absolute error between the load prediction results and the actual load results of any energy type, and the error coefficient of the energy type is obtained according to the absolute error as:

[0037] where CE p represents the error coefficient of the p-th energy type, Upred p,t represents the predicted data of the t-th time node of the p-th energy type, Uactual p,t represents the actual data of the t-th time node of the p-th energy type, R represents that there are R time node data in the prediction results, and ln() represents the logarithmic function, represents the predicted mean value of the R time nodes of the p-th energy type, represents the actual mean value of the R time nodes of the p-th energy type.

[0038] On the other hand, based on the control algorithm, combining the error coefficient of the energy type and the standard power generation data, the acquisition frequency of the energy is adjusted, including:

[0039] Taking the error coefficient of any energy type as the proportional coefficient of the control algorithm, the standard power generation data as the differential comparison coefficient, and the error accumulation as the integral coefficient;

[0040] Inputting the proportional coefficient, the differential comparison coefficient, and the integral coefficient into the control algorithm to output the acquisition frequency adjustment signal of the energy type. Among them, if the error coefficient is greater than the preset standard threshold, the proportional coefficient is a positive number; if the error coefficient is less than or equal to the preset standard threshold, the proportional coefficient is a negative number;

[0041] Sending the acquisition frequency adjustment signal to the corresponding controller according to the energy type, and carrying out real-time adjustment on the acquisition of the energy type to obtain the frequency control result.

[0042] On the other hand, evaluating the frequency control result, and reversely optimizing the control algorithm and the load prediction model based on the evaluation result, including:

[0043] Configuring the load prediction model and the control algorithm into the actual environment, obtaining the real-time power generation data after control adjustment of any energy type, generating the real-time prediction data corresponding to the real-time power generation data time period according to the control algorithm, and combining the real-time power generation data and the real-time prediction data to obtain the frequency adjustment error;

[0044] Reverse-optimize the control algorithm and load prediction model based on the frequency adjustment error, and adjust the model parameters and the weight coefficients of the control algorithm.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] The present invention provides a method for controlling the data acquisition frequency of multiple energy sources in a power grid considering load prediction errors, which is used to improve the accuracy of power grid dispatching by gradually optimizing the data acquisition and load prediction processes of the multiple energy source system in the power grid. First, determine different energy types and standardize the original power generation data; then, build a load prediction model, analyze the errors and obtain the error coefficients; next, adjust the acquisition frequency through a control algorithm to optimize the data acquisition process; finally, reverse-optimize the control algorithm and the prediction model based on the evaluation results to ensure the efficiency and stability of the system. This method can effectively improve the accuracy and operability of the multiple energy source dispatching in the power grid. Description of the Drawings

[0047] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a schematic flowchart of the method for controlling the data acquisition frequency of multiple energy sources in a power grid considering load prediction errors provided by an embodiment of the present invention. Detailed Embodiments

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions in the present invention with reference to the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0050] Embodiment 1:

[0051] As Figure 1 shown, the method for controlling the data acquisition frequency of multiple energy sources in a power grid considering load prediction errors provided by an embodiment of the present invention includes:

[0052] Step 1: Determine the energy types of the multiple energy sources in the studied power grid, collect the original data of the energy according to the energy characteristics of any one of the energy types, and standardize the original data to obtain standard power generation data;

[0053] Step 2: Build a load forecasting model, using the standard power generation data of all energy types as the input of the load forecasting model, and output the load forecasting results;

[0054] Step 3: Conduct error analysis on the load forecasting results and the actual load results to obtain the error coefficients for the corresponding energy types;

[0055] Step 4: Based on the control algorithm, combine the error coefficients of the energy types and the standard power generation data to adjust the acquisition frequency of the energy;

[0056] Step 5: Evaluate the frequency control results, and reversely optimize the control algorithm and the load forecasting model based on the evaluation results.

[0057] In this embodiment, the power grid is a complex system for transmitting and distributing electric power, which transmits the electric power generated by power plants to end-users through facilities such as transmission lines, substations, and distribution networks.

[0058] In this embodiment, the energy types refer to energy from different sources, such as machine gas, natural gas, petroleum, etc.

[0059] In this embodiment, the energy characteristics refer to the characteristics and behaviors of the energy during the power generation process, including the way it generates electricity, stability, dispatchability, and dependence on the external environment, etc.

[0060] In this embodiment, the raw data refers to the unprocessed basic data directly collected from energy production measurement devices.

[0061] In this embodiment, standardization is a common method in data preprocessing, which converts raw data into a unified scale or range for subsequent analysis and modeling.

[0062] In this embodiment, the standard power generation data refers to the power generation data of various energy types after being standardized.

[0063] In this embodiment, the load forecasting model is a model used to predict the power demand (i.e., load) in the power system.

[0064] In this embodiment, the load forecasting results refer to the predicted values of power demand output by the load forecasting model, which refer to the power load demand in a future certain time period.

[0065] In this embodiment, the actual load results refer to the actual power demand occurring in the power grid during a specific time period.

[0066] In this embodiment, the error coefficient refers to a measure of the difference between the load forecasting results and the actual load results during the load forecasting process.

[0067] In this embodiment, the control algorithm refers to a mathematical method used to regulate and optimize processes such as energy harvesting, load scheduling, and energy distribution.

[0068] In this embodiment, the acquisition frequency refers to the time interval of data acquisition or the frequency of acquisition.

[0069] In this embodiment, the frequency control result refers to the actual output and performance of the energy data acquisition frequency after being adjusted by the control algorithm.

[0070] The working principle and beneficial effects of the above technical solution are as follows: By standardizing energy data, building a load prediction model, error analysis, and frequency control, the data acquisition and scheduling process of the multi - energy system of the power grid is optimized. By dynamically adjusting the acquisition frequency, the prediction error is reduced, and the operation efficiency and stability of the power grid are improved.

[0071] Embodiment 2:

[0072] Based on the above Embodiment 1, determine the energy types of the multi - energy of the power grid to be studied, and collect the original data of the energy according to the energy characteristics of any energy type, including:

[0073] According to the configuration information of the target power grid to be studied, obtain all the energy types connected to the target power grid to be studied;

[0074] Select the corresponding sensor type according to the energy type, and identify the installation locations of all sensors according to the three - dimensional geographical map of the target power grid to be studied;

[0075] Based on the energy characteristics of any energy type, obtain the matching scores of the corresponding sensors and any installation location, select the installation location with the highest matching score of the sensor as the optimal installation location, configure a unique first number for the optimal installation location, and configure a unique second number for the sensor;

[0076] According to the corresponding relationship between the first number and the second number, complete the installation and configuration of all sensors, and collect the original data of the corresponding energy type based on the sensors.

[0077] In this embodiment, the configuration information includes: power grid structure, power grid capacity, load conditions, energy types, etc.

[0078] In this embodiment, the sensor types include: electric energy, wind energy, temperature, humidity, etc.

[0079] In this embodiment, the three - dimensional geographical map refers to a technology that represents the earth's surface and related objects, structures, and geographical information in the geographical space using a three - dimensional coordinate system.

[0080] In this embodiment, the installation location refers to the optimal geographical coordinates selected for a specific type of sensor based on the three-dimensional geographical map of the target research power grid.

[0081] In this embodiment, the matching score is to evaluate the degree of fit between the working performance of a certain sensor at a specific installation location and the energy harvesting conditions that the location can provide.

[0082] In this embodiment, the optimal installation location refers to the best location selected by calculating the matching score according to the configuration information of the target research power grid, the characteristics of the energy type, the sensor type, and the geographical location characteristics.

[0083] In this embodiment, the first number refers to the unique identifier of the optimal installation location.

[0084] In this embodiment, the second number refers to the unique identifier of the sensor.

[0085] The working principle and beneficial effects of the above technical solution are as follows: By analyzing the power grid configuration information, selecting matching sensors and determining the optimal installation location, the best configuration of the sensors and the data acquisition effect are ensured. Through number management, the installation process is optimized, and the accuracy and efficiency of data acquisition are improved.

[0086] Embodiment 3:

[0087] Based on the above Embodiment 1, the original data is standardized to obtain standard power generation data, including:

[0088] Obtain the original data of all energy types, perform time alignment processing on the original data of any energy type according to a preset standard time interval to obtain the first power generation data;

[0089] Identify the missing values in the first power generation data, and fill the missing values based on zero mean and unit variance to obtain the standard power generation data;

[0090] Initialize the power generation database according to the preset field type, use the energy type and the collection time as the combined primary key, and store the standard power generation data in the power generation database.

[0091] In this embodiment, the preset standard time interval refers to the fixed time interval used for time alignment processing of the original data of the energy type.

[0092] In this embodiment, the time alignment processing refers to adjusting the energy data from different sources and different timestamps according to a unified time interval so that they are aligned in the time dimension.

[0093] In this embodiment, the first power generation data refers to the data set obtained after time alignment processing, in which the data of all energy types are sorted and aligned according to a unified time interval.

[0094] In this embodiment, a missing value refers to a situation in a dataset where some data items are not collected or are not correctly recorded, resulting in no valid data at that position.

[0095] In this embodiment, zero-mean unit variance is a method of data standardization used to transform data into a standard normal distribution with a mean of 0 and a variance of 1.

[0096] In this embodiment, standard power generation data refers to power generation data that has been time-aligned and standardized.

[0097] In this embodiment, initialization is the process of setting up, configuring, or preparing a system, data structure, or resource.

[0098] In this embodiment, a power generation database is a database used to store power generation data of different energy types (such as wind energy, solar energy, thermal power, hydropower, etc.).

[0099] The working principle and beneficial effects of the above technical solution are as follows: Through time alignment processing and missing value filling, the integrity and standardization of energy data are ensured. Through database initialization and storage, efficient management and query of energy data are achieved, improving the accuracy and reliability of data processing.

[0100] Example 4:

[0101] Based on the above Example 3, before building a load forecasting model, use feature engineering to process the standard power generation data to obtain energy power generation features, including:

[0102] Based on the power generation database, obtain the standard power generation data of all energy types, and use feature engineering to process the standard power generation data to obtain energy power generation features;

[0103] According to the standard power generation data, construct a feature matrix L:

[0104] where a ij represents the data of the j-th time node of the i-th energy type, n represents that each energy type has n time nodes, and m represents that there are m energy types in total;

[0105] Construct a divergence matrix B as: where μ k represents the mean vector of the k-th energy type, represents the mean of all standard power generation data, T represents the transpose symbol, and · represents the dot product symbol;

[0106] According to the feature matrix L and the divergence matrix B, and using a maximization function to process, obtain a projection feature matrix W:

[0107] Maximize the variance of the between-class scatter matrix B in the low-dimensional space, and at the same time minimize the variance of the within-class feature matrix L in the low-dimensional space to obtain the optimal projection feature matrix W;

[0108] According to L = W·ω, the eigenvector is obtained

[0109] Among them, ω1 represents the power generation feature of the first energy type, ω2 represents the power generation feature of the second energy type, and ω m represents the power generation feature of the m-th energy type.

[0110] In this embodiment, feature engineering is a very important process in data science and machine learning, which involves extracting, transforming, constructing, and selecting useful features from raw data.

[0111] In this embodiment, the power generation feature is a representative and discriminative feature extracted by processing power generation data through feature engineering.

[0112] In this embodiment, the feature matrix is a matrix constructed by key features extracted from standard power generation data and is used to describe the power generation data of different energy types.

[0113] In this embodiment, the scatter matrix is a matrix used to measure the relationship and degree of variation between variables in a data set.

[0114] In this embodiment, the maximization function is a function that obtains a projection matrix through optimization so that the between-class scatter is as large as possible in the low-dimensional space.

[0115] In this embodiment, the optimal projection feature matrix is a matrix obtained to represent the maximization of the projection between-class scatter.

[0116] The working principle and beneficial effects of the above technical solution are: by processing power generation data through feature engineering, constructing a feature matrix and a scatter matrix, and using the optimization method of maximizing the between-class scatter and minimizing the within-class scatter, the optimal projection feature matrix is obtained, realizing the effective extraction and dimensionality reduction of power generation features, and improving the efficiency and accuracy of data analysis.

[0117] Embodiment 5:

[0118] On the basis of the above Embodiment 4, a load prediction model is built. The standard power generation data of all energy types are used as the input of the load prediction model, and the load prediction results are output, including:

[0119] Build the framework of the load prediction model according to the preset prediction model, obtain the historical power generation data by matching the data in the historical library according to the energy type, use the historical power generation data as the training set, and use the power generation feature as the validation set;

[0120] Iteratively train a preset prediction model based on a training set to obtain first model parameters, and use the first model parameters as framework nodes to obtain a first load prediction model;

[0121] Validate the first load prediction model based on a validation set to obtain model hyperparameters, and optimize and adjust the first model parameters according to the Bayesian optimization algorithm combined with the hyperparameters to obtain final model parameters;

[0122] Input the final model parameters into the first load prediction model to obtain a final load prediction model;

[0123] Use the standard power generation data of all energy types as the input of the load prediction model and output the load prediction result.

[0124] In this embodiment, the preset prediction model refers to the prediction model initially selected in the load prediction system and is the starting point of the entire framework.

[0125] In this embodiment, the historical database refers to the database for storing and managing historical data.

[0126] In this embodiment, the historical power generation data refers to the data records of the power generation of different energy types (such as coal power, gas power, wind energy, solar energy, etc.) over a past period of time.

[0127] In this embodiment, the training set refers to the data set used to train the prediction model.

[0128] In this embodiment, the validation set is a part of the data divided from the original data and is used to evaluate the performance of the model during the training process.

[0129] In this embodiment, the first model parameters refer to the model parameters obtained by using the training set (historical power generation data) in the preliminary training stage of the preset prediction model, including: model weights, initial hyperparameters, activation function parameters, etc.

[0130] In this embodiment, the framework node refers to a node that is a key component or intermediate step in the model construction process, and this node contains the core parameters and structure of the model.

[0131] In this embodiment, the first load prediction model is a preliminary load prediction model obtained after preliminary training based on the training set and the initially set preset prediction model.

[0132] In this embodiment, the model hyperparameters refer to the parameters that need to be manually set before training the model, including: learning rate, number of iterations, regularization coefficient.

[0133] In this embodiment, the Bayesian optimization algorithm is an optimization algorithm for optimizing model parameters. Through a surrogate model, it fits the objective function based on known data points and gives the uncertainty of the function value (i.e., the variance of the prediction).

[0134] In this embodiment, the final model parameters refer to the optimized parameters obtained through the following process.

[0135] In this embodiment, the final load prediction model is a high-precision prediction tool obtained by optimizing the initial model parameters, adjusting the model hyperparameters, and performing iterative training.

[0136] The working principle and beneficial effects of the above technical solution are as follows: The load prediction model is trained with historical power generation data, and the Bayesian optimization algorithm is combined to adjust the model parameters to improve the prediction accuracy. Finally, the optimized model is used for load prediction to improve the accuracy and reliability of energy load prediction.

[0137] Embodiment 6:

[0138] Based on the above Embodiment 1, the error analysis is performed on the load prediction result and the actual load result to obtain the error coefficient corresponding to the energy type, including:

[0139] Perform error analysis on the load prediction result and the actual load result of any energy type to obtain the absolute error between the load prediction result and the actual load result of any energy type, and obtain the error coefficient of the energy type according to the absolute error as:

[0140] where CE p represents the error coefficient of the p-th energy type, Upred p,t represents the predicted data of the t-th time node of the p-th energy type, Uactual p,t represents the actual data of the t-th time node of the p-th energy type, R represents that there are R time node data in the prediction result, ln() represents the logarithmic function, represents the predicted mean of R time nodes of the p-th energy type, represents the actual mean of R time nodes of the p-th energy type.

[0141] In this embodiment, the absolute error refers to the absolute value of the difference between the predicted value and the actual value.

[0142] The working principle and beneficial effects of the above technical solution are as follows: By comparing the load prediction result with the actual data, calculating the absolute error and introducing the error coefficient, and normalizing the error using the logarithmic function, the prediction accuracy of the energy type is evaluated. This method improves the accuracy of error analysis and helps to optimize the prediction model.

[0143] Example 7:

[0144] Based on the above Example 6, based on the control algorithm, combined with the error coefficient of the energy type and the standard power generation data, the acquisition frequency of the energy is adjusted, including:

[0145] Take the error coefficient of any energy type as the proportional coefficient of the control algorithm, the standard power generation data as the differential comparison coefficient, and the error accumulation as the integral coefficient;

[0146] Input the proportional coefficient, differential comparison coefficient, and integral coefficient into the control algorithm, and output the acquisition frequency adjustment signal of the energy type. Among them, if the error coefficient is greater than the preset standard threshold, the proportional coefficient is positive; if the error coefficient is less than or equal to the preset standard threshold, the proportional coefficient is negative;

[0147] Send the acquisition frequency adjustment signal to the corresponding controller according to the energy type, and perform real-time adjustment on the acquisition of the energy type to obtain the frequency control result.

[0148] In this embodiment, the proportional coefficient is a parameter used to adjust the acquisition frequency adjustment signal and plays a role in determining the response intensity.

[0149] In this embodiment, the differential comparison coefficient is a parameter used to reflect the change rate of the system error. It is related to differential control, used to calculate the change rate of the error, and add a response to the error change trend to the control signal.

[0150] In this embodiment, the integral coefficient refers to the coefficient related to the accumulation of errors, used to reflect the accumulation effect of errors over a period of time, and adjust the control signal through the historical accumulation of error values, aiming to eliminate long-term accumulated errors.

[0151] In this embodiment, the frequency adjustment signal is a control signal generated by the combined action of the proportional coefficient, differential comparison coefficient, and integral coefficient.

[0152] In this embodiment, the preset standard threshold is a critical value used to judge the error coefficient, which determines when the sign of the proportional coefficient will change.

[0153] In this embodiment, the controller is a device used to process and adjust the acquisition frequency of the energy type.

[0154] The working principle and beneficial effects of the above technical solution are: by taking the error coefficient as the proportional coefficient of the control algorithm, combining differential and integral coefficients to adjust the acquisition frequency of the energy type, and optimizing the data acquisition frequency in real time. By dynamically adjusting the frequency, the system response speed and acquisition accuracy are improved, and the energy management effect is optimized.

[0155] Example 8:

[0156] Based on the above-mentioned Embodiment 1, the frequency control result is evaluated, and the control algorithm and the load prediction model are reversely optimized based on the evaluation result, including:

[0157] Configure the load prediction model and the control algorithm into the actual environment, obtain the real-time power generation data after control and regulation of any energy type, generate real-time prediction data for the corresponding real-time power generation data time period according to the control algorithm, and obtain the frequency adjustment error by combining the real-time power generation data and the real-time prediction data;

[0158] Based on the frequency adjustment error, reversely optimize the control algorithm and the load prediction model, and adjust the model parameters and the weight coefficients of the control algorithm.

[0159] In this embodiment, the actual environment refers to the real operation environment or actual application scenario for load prediction, control algorithm execution, and energy acquisition and regulation.

[0160] In this embodiment, the real-time power generation data refers to various data that reflect the power generation output of a certain energy production facility (such as a generator set, a solar panel, a wind turbine, etc.) in real time at a specific time point or time period.

[0161] In this embodiment, the real-time prediction data refers to the prediction result of the future power load for a certain period of time based on the existing data and model within a given time period.

[0162] In this embodiment, the frequency adjustment error refers to the difference between the actual power generation and the power generation predicted by the load prediction model or the control algorithm in the power system.

[0163] The working principle and beneficial effects of the above technical solution are: by obtaining the power generation data and prediction data in real time, calculating the frequency adjustment error, reversely optimizing based on the error, adjusting the load prediction model and control algorithm parameters, improving the accuracy and response ability of the system, realizing dynamic optimization and intelligent regulation, and optimizing the energy dispatch efficiency.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecasting errors, characterized in that: include: Step 1: Determine the energy type of the multi-energy source of the research power grid, and collect the original data of the energy according to the energy characteristics of any energy type, and standardize the original data to obtain standard power generation data; Step 2: Build a load forecasting model, use the standard power generation data of all energy types as the input of the load forecasting model, and output the load forecasting results; Step 3: Perform error analysis on the load forecast results and the actual load results to obtain the error coefficient of the corresponding energy type; Step 4: Based on the control algorithm, combined with the error coefficient of the energy type and the standard power generation data, adjust the energy collection frequency; Step 5: Evaluate the frequency control results and reversely optimize the control algorithm and load forecasting model based on the evaluation results.

2. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 1 is characterized in that: Determine the energy type of the multi-energy source of the research grid, and collect the original data of the energy according to the energy characteristics of any energy type, including: According to the configuration information of the target research power grid, all energy types connected to the target research power grid are obtained; Select the corresponding sensor type according to the energy type, and identify the installation locations of all sensors based on the three-dimensional geographic map of the target research power grid; Based on the energy characteristics of any energy type, obtain a matching score between the corresponding sensor and any installation position, select the installation position of the sensor with the highest matching score as the optimal installation position, configure a unique first number for the optimal installation position, and configure a unique second number for the sensor; According to the corresponding relationship between the first number and the second number, all sensors are installed and configured, and raw data of corresponding energy types are collected based on the sensors.

3. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 1 is characterized in that: Standardized raw data is used to obtain standard power generation data, including: Acquire raw data of all energy types, perform time alignment processing on the raw data of any energy type according to a preset standard time interval, and obtain first power generation data; Identify missing values ​​of the first power generation data, fill the missing values ​​based on zero mean and unit variance, and obtain standard power generation data; The power generation database is initialized according to the preset field type, the energy type and the collection time are used as the joint primary key, and the standard power generation data is stored in the power generation database.

4. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 3 is characterized in that: Before building the load forecasting model, feature engineering is used to process standard power generation data to obtain energy generation characteristics, including: Based on the power generation database, standard power generation data of all energy types are obtained, and feature engineering is used to process the standard power generation data to obtain energy power generation characteristics; According to the standard power generation data, construct the characteristic matrix L: Among them, a ij Represents the data of the jth time node of the i-th energy type, n means that each energy type has n time nodes, and m means that there are m energy types in total; Construct the divergence matrix B as: Among them, μ k represents the mean vector of the k-th energy type, represents the mean of all standard power generation data, T represents the transposition symbol, and · represents the dot product symbol; According to the feature matrix L and the divergence matrix B, and using the maximization function, the projection feature matrix W is obtained: Maximize the variance of the inter-class scatter matrix B in the low-dimensional space, and minimize the variance of the intra-class feature matrix L in the low-dimensional space to obtain the optimal projection feature matrix W; According to L = W·ω, we get the eigenvector Among them, ω1 represents the energy generation characteristics of the first energy type, ω2 represents the energy generation characteristics of the second energy type, and ω m Represents the energy generation characteristics of the mth energy type.

5. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 4 is characterized in that: Build a load forecasting model, use the standard power generation data of all energy types as the input of the load forecasting model, and output the load forecasting results, including: The framework of the load forecasting model is built according to the preset forecasting model. The historical power generation data is obtained by matching the data in the historical database according to the energy type. The historical power generation data is used as the training set, and the energy generation characteristics are used as the verification set. Iteratively train the preset prediction model based on the training set to obtain first model parameters, and use the first model parameters as framework nodes to obtain a first load prediction model; The first load forecasting model is verified based on the verification set to obtain model hyperparameters, and the first model parameters are optimized and adjusted according to the Bayesian optimization algorithm combined with the hyperparameters to obtain the final model parameters; Inputting the final model parameters into the first load forecasting model to obtain a final load forecasting model; The standard power generation data of all energy types is used as the input of the load forecasting model to output the load forecast results.

6. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 1, characterized in that: Perform error analysis on the load forecast results and the actual load results to obtain the error coefficient of the corresponding energy type, including: The load forecast result and the actual load result of any energy type are subjected to error analysis to obtain the absolute error between the load forecast result and the actual load result of any energy type, and the error coefficient of the energy type is obtained according to the absolute error: Among them, CE p represents the error coefficient of the p-th energy type, Upred p,t represents the forecast data of the p-th energy type at the t-th time node, Uactual p,t represents the actual data of the tth time node of the pth energy type, R represents the predicted result with data of R time nodes in total, ln() represents the logarithmic function, represents the predicted mean value of the p-th energy type at R time nodes, Represents the actual mean value of the p-th energy type at R time nodes.

7. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 6, characterized in that: Based on the control algorithm, combined with the error coefficient of the energy type and the standard power generation data, the energy collection frequency is adjusted, including: The error coefficient of any energy type is used as the proportional coefficient of the control algorithm, the standard power generation data is used as the differential comparison coefficient, and the error accumulation is used as the integral coefficient; Input the proportional coefficient, differential contrast coefficient, and integral coefficient into the control algorithm, and output the acquisition frequency adjustment signal of the energy type, wherein if the error coefficient is greater than the preset standard threshold, the proportional coefficient is a positive number, and if the error coefficient is less than or equal to the preset standard threshold, the proportional coefficient is a negative number; The acquisition frequency adjustment signal is transmitted to the corresponding controller according to the energy type, and the acquisition of the energy type is adjusted in real time to obtain a frequency control result.

8. The method for controlling the frequency of multi-energy data acquisition in a power grid considering load forecast error according to claim 1, characterized in that: Evaluate the frequency control results and reversely optimize the control algorithm and load forecasting model based on the evaluation results, including: The load forecasting model and control algorithm are configured in the actual environment, the real-time power generation data after control and adjustment of any energy type is obtained, the real-time forecasting data corresponding to the time period of the real-time power generation data is generated according to the control algorithm, and the frequency adjustment error is obtained by combining the real-time power generation data and the real-time forecasting data; Based on the frequency adjustment error, the control algorithm and load forecasting model are reversely optimized to adjust the model parameters and the weight coefficients of the control algorithm.

Images