Supply prediction system for renewable energy sources in power system

Abstract

The invention relates to a renewable energy supply prediction system in a power system in the technical field of energy supply, and the system comprises a data obtaining and preprocessing module which is used for obtaining the energy purchase cost and supply amount of a renewable energy supplier at a target moment, and the data of the energy load attribute of a user; the feature extraction and selection module is used for extracting key features remarkably influencing energy supply from the data by using mutual information entropy and principal component analysis technologies; according to the invention, two advanced machine learning algorithms including LightGBM and XGBoost are combined, so that the system can predict the supply condition of renewable energy more accurately. The initial prediction uses a LightGBM model to extract an initial result, and then the secondary prediction uses an XGBoost algorithm to perform further optimization based on the results and key features, thereby improving the accuracy of the overall prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of energy supply, and in particular to a supply prediction system for renewable energy in a power system. Background Art

[0002] With the global emphasis on environmental protection and sustainable development, the development of renewable energy (such as solar energy, wind energy, etc.) has become an important force in the international energy transformation. However, renewable energy has uncertain characteristics such as volatility and intermittency, which pose challenges to the operation and dispatching of the power system. Especially in the context of high proportion of renewable energy access, how to accurately predict its output and effectively quantify the uncertainty is of great significance for optimizing the operation and dispatching of wind power and photovoltaic power plants and ensuring the safe and stable operation of the power system.

[0003] 1. Status Quo and Challenges of Renewable Energy Generation 1.1 Characteristics of Renewable Energy Volatility and intermittency: For example, photovoltaic power generation depends on sunlight intensity, while wind power generation depends on wind speed. The changes in these natural factors lead to the volatility and intermittency of power generation.

[0004] Regional differences: There are significant differences in light conditions and wind speed distributions in different regions, resulting in different renewable energy generation potentials and actual outputs in each region.

[0005] 1.2 Existing Problems Wind and light curtailment phenomena: Due to the volatility and intermittency of renewable energy, the power grid is difficult to fully absorb, resulting in waste of some electricity.

[0006] Decline in power system stability: The inertia of the highly power-electronic power system decreases, and its stability is lower than that of the traditional power system, further increasing the dispatching difficulty.

[0007] Limitations of subjective evaluation: At present, the suitability evaluation of renewable energy power plant siting and construction projects based on counties mainly refers to factors such as regional energy demand, economic development, and power generation potential, and has a certain degree of subjectivity. There is an urgent need for a new breakthrough in research methods.

[0008] In order to address the above problems, it is particularly important to develop a supply prediction system for renewable energy in a power system that integrates functions such as data acquisition, preprocessing, feature extraction and selection, primary prediction, and secondary prediction. The system should have the following characteristics: a multi-step prediction mechanism, advanced data analysis techniques, and an optimized energy supply strategy to improve the utilization efficiency of renewable energy.

[0009] In summary, the present invention aims to provide a supply prediction system for renewable energy in a power system. Summary of the Invention

[0010] In order to overcome the deficiencies in the background art, the present invention discloses a supply prediction system for renewable energy in a power system.

[0011] To achieve the above-mentioned invention purpose, the present invention adopts the following technical solutions: A supply prediction system for renewable energy in a power system, comprising: A data acquisition and preprocessing module, configured to acquire data on the energy procurement cost, supply volume of the renewable energy supplier at the target moment, and the energy load attributes of users; A feature extraction and selection module, configured to extract key features that significantly affect energy supply from the above data by using mutual information entropy and principal component analysis (PCA) techniques; A primary prediction module, configured to construct a LightGBM model, perform a primary prediction based on the above key features, and use the prediction result as a new variable; A secondary prediction module, configured to use the result of the primary prediction and the key features as input values, and perform a secondary prediction using the XGBoost algorithm to obtain the final prediction result; A target energy supply information acquisition module, configured to determine the optimal energy supply strategy related to users according to the above final prediction result.

[0012] Preferably, the feature extraction and selection module has an outlier processing unit for identifying and removing or replacing outliers in the original data.

[0013] Preferably, the primary prediction module has a parameter update unit for continuously adjusting and optimizing the parameters of the LightGBM model according to the prediction error.

[0014] Preferably, the secondary prediction module has an objective function construction unit for constructing an objective function including a loss function and a regularization term, and optimizing by expanding the loss function using the Taylor formula.

[0015] A method for predicting the supply of renewable energy in a power system includes the following steps: S1: A data acquisition step for acquiring the energy procurement cost, supply volume of the renewable energy supplier at the target moment, and the energy load attributes of users; S2: A data preprocessing step for performing exploratory analysis on the acquired data and processing outliers to make the data continuous and periodic; S3: A key feature confirmation step for determining key features according to the field of use; S4: A feature variable extraction step for extracting key feature variables that significantly affect energy supply from the above data by using mutual information entropy and principal component analysis (PCA) techniques; S5: Initial prediction step, constructing a LightGBM model, using the extracted key features as input values for initial prediction, and taking the prediction result as a new variable; S6: Secondary prediction step, using the result of the initial prediction and the key features as input values, and applying the XGBoost algorithm for secondary prediction to obtain the final prediction result; S7: Target energy supply information acquisition step, determining the optimal energy supply strategy related to the user according to the final prediction result.

[0016] Preferably, in S2, the power generation of renewable energy is regarded as a continuous variable in a cycle, and the dataset is explored and analyzed according to the time sequence and periodic law, and a scatter plot of the characteristic variables is drawn to eliminate outliers and extreme points to ensure the continuity and periodicity of the data.

[0017] Preferably, in S4, the information entropy describes the degree of uncertainty of the values taken by a random variable in a numerical form and quantitatively describes the information content of the random variable.

[0018] Preferably, in S5, GOSS sorts according to the absolute value of the gradients of all samples, selects the first a * 100% samples as set A, and the remaining (1 - a) * 100 samples as set AC; Randomly select b * 100 samples as set B, where the size of B is b|AC|, to generate a small-gradient sample set; Multiply a constant coefficient when calculating the information gain of the small-gradient samples, and perform the initial prediction based on the LightGBM model, taking the prediction result as a new variable.

[0019] Preferably, in S6, the sample set xi (i = 1, 2,..., 1974) is identified, and a regression tree is constructed; Construct an objective function, including a loss function and a regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model; Expand the loss function using the Taylor formula to obtain the final objective function, and perform secondary prediction based on the XGBoost algorithm to obtain the final prediction result.

[0020] Due to the adoption of the above-mentioned technical solution, the present invention has the following beneficial effects: 1. By combining two advanced machine learning algorithms, LightGBM and XGBoost, the system can more accurately predict the supply situation of renewable energy. The initial prediction uses the LightGBM model to extract preliminary results, and then the secondary prediction is further optimized based on these results and key features using the XGBoost algorithm, thereby improving the overall prediction accuracy; 2. A multi-step prediction mechanism is designed, and mutual information entropy and principal component analysis (PCA) techniques are used to extract key features that significantly affect energy supply, which helps to improve the generalization ability of the model in different scenarios and reduce the occurrence of overfitting and underfitting phenomena; 3. Determine the optimal energy supply strategy related to users according to the final prediction result, which helps to optimize energy supply, reduce waste, and thus improve the energy utilization efficiency of the entire power system; 4. By accurately predicting the supply volume of renewable energy, power resources can be better dispatched and managed, the grid instability problems caused by the volatility and intermittency of renewable energy can be alleviated, and the stability and security of the power system can be improved; 5. This system not only provides data prediction functions, but also provides a scientific basis for formulating energy supply strategies by comprehensively considering various factors such as energy procurement costs, supply volumes, and user energy load attributes, helping relevant decision-makers make more informed choices. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flowchart of the present invention; Figure 2 is a flowchart of data acquisition and preprocessing in the present invention; Figure 3 is a flowchart of feature extraction and selection in the present invention; Figure 4 is a flowchart of the initial prediction in the present invention; Figure 5 is a flowchart of the secondary prediction in the present invention.

[0022] In the figure: 100, data acquisition and preprocessing module; 200, feature extraction and selection module; 300, initial prediction module; 400, secondary prediction module; 500, target energy supply information acquisition module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The present invention can be explained in detail through the following embodiments. The purpose of disclosing the present invention is to protect all technical improvements within the scope of the present invention. In the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "front", "rear", "left", "right", etc. indicating the orientation or positional relationship, it is only corresponding to the drawings of the present application for the convenience of describing the present invention; it should be understood that if there are terms such as "end", "side", "end part", "side part", "lateral", "longitudinal", etc. indicating the orientation or positional relationship, it is only corresponding to the length and width of the corresponding component, that is, the "end part" indicates the head and tail regions in the length direction of the corresponding component, and the "side part" indicates the head and tail regions in the width direction of the corresponding component; it is for the convenience of describing the present invention rather than indicating or implying that the device or element referred to must have a specific orientation.

[0024] Example 1, in combination with the attached Figures 1-5 , for an application in the field of photovoltaic power generation, a supply prediction system for renewable energy in a power system, including: Data acquisition and preprocessing module 100: Obtain the energy procurement cost, supply volume, and the energy load attributes of users within a specific time period from the area where a certain photovoltaic power station is located; according to requirements, the energy load attributes include temperature demand attributes and balance attributes.

[0025] Conduct an exploratory analysis on the acquired data, conduct a preliminary analysis on the data set according to the time sequence and periodic law, and draw a scatter plot of characteristic variables to identify outliers. For the discovered outliers, the outlier processing unit identifies and eliminates or replaces the outliers in the original data, for example, using the previous value in the same column for replacement, to ensure the continuity and periodicity of the data.

[0026] Feature extraction and selection module 200: Utilize mutual information entropy and principal component analysis (PCA) techniques to extract key characteristic variables that significantly affect the supply of photovoltaic electric energy from the original data, such as sunshine duration, temperature change, etc.

[0027] Initial prediction module 300: Construct a LightGBM model and use the extracted key features as input values for initial prediction. Specifically, first sort according to the absolute value of the gradient of all samples, select the first a*100% samples as set A, and the remaining (1 - a)100 samples as set AC. Randomly select b100 samples as set B to generate a small gradient sample set, and conduct an initial prediction based on the LightGBM model to obtain a preliminary prediction result.

[0028] According to requirements, the initial prediction module 300 has a parameter update unit for continuously adjusting and optimizing the parameters of the LightGBM model according to the prediction error.

[0029] Secondary prediction module 400: Use the results of the initial prediction and the key features as input values and apply the XGBoost algorithm for secondary prediction. Construct an objective function through its internal objective function construction unit, including a loss function and a regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model. Expand the loss function using the Taylor formula to obtain the final objective function, and conduct secondary prediction based on the XGBoost algorithm to obtain the final prediction result.

[0030] Target energy supply information acquisition module 500: Determine the best energy supply strategy related to the user according to the final prediction result, optimize the energy supply, reduce waste, and improve energy utilization efficiency.

[0031] Specifically, it includes the following steps: Step 1: Data acquisition and preprocessing Obtain the energy procurement cost, supply volume in a specific time period in the area where a certain photovoltaic power station is located, and the energy load attributes of users (including temperature demand attributes and balance attributes).

[0032] Conduct an exploratory analysis on the acquired data, conduct a preliminary analysis on the data set according to the time sequence and periodic law, and draw a scatter plot of characteristic variables to identify outliers. For the discovered outliers, replace them with the previous value in the same column to ensure the continuity and periodicity of the data.

[0033] Step 2: Confirmation of key features Analyze the working principle and power input-output characteristics of the photovoltaic power generation system, and determine the key factors affecting the photovoltaic power supply, such as sunlight intensity, ambient temperature, etc.

[0034] Step 3: Extraction of characteristic variables Use mutual information entropy and principal component analysis (PCA) techniques to extract key characteristic variables that significantly affect the photovoltaic power supply from the original data. For example, sunshine duration, temperature change, etc.

[0035] Step 4: Initial prediction Build a LightGBM model and use the extracted key features as input values for the initial prediction. Specifically, first sort according to the absolute value of the gradient of all samples, select the first a*100% samples as set A, and the remaining (1 - a)100 samples as set AC. Randomly select b100 samples as set B to generate a small gradient sample set, and conduct an initial prediction based on the LightGBM model to obtain a preliminary prediction result.

[0036] Step 5: Secondary prediction Use the result of the initial prediction and the key features as input values, and apply the XGBoost algorithm for secondary prediction. Build an objective function, including a loss function and a regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model. Expand the loss function using Taylor's formula to obtain the final objective function, and conduct secondary prediction based on the XGBoost algorithm to obtain the final prediction result.

[0037] Step 6: Obtaining target energy supply information Determine the best energy supply strategy related to the user according to the final prediction result, optimize the energy supply, reduce waste, and improve energy utilization efficiency.

[0038] Example 2, in combination with the attached Figures 1-5 , in the application in the field of wind power generation, a supply prediction system for renewable energy in a power system, including: Data acquisition and preprocessing module 100: Obtain the energy procurement cost, supply volume, and the energy load attributes of users in a specific time period from the area where a wind farm is located. As needed, the energy load attributes include temperature demand attributes and balance attributes).

[0039] Conduct exploratory analysis on the acquired data, perform preliminary analysis on the data set according to the time sequence and periodic law, and draw a scatter plot of characteristic variables to identify outliers. For the discovered outliers, the outlier processing unit identifies and removes or replaces the outliers in the original data, for example, using the previous value in the same column for replacement, to ensure the continuity and periodicity of the data.

[0040] Feature extraction and selection module 200: Utilize mutual information entropy and principal component analysis (PCA) techniques to extract key characteristic variables that significantly affect wind power supply from the original data, such as average wind speed, maximum wind speed, etc.

[0041] Initial prediction module 300: Construct a LightGBM model and use the extracted key features as input values for initial prediction. Specifically, first sort according to the absolute value of the gradient of all samples, select the first a*100% samples as set A, and the remaining (1 - a)100 samples as set AC. Randomly select b100 samples as set B to generate a small gradient sample set, and perform initial prediction based on the LightGBM model to obtain a preliminary prediction result.

[0042] As needed, the initial prediction module 300 has a parameter update unit for continuously adjusting and optimizing the parameters of the LightGBM model according to the prediction error.

[0043] Secondary prediction module 400: Use the results of the initial prediction and the key features as input values and perform secondary prediction using the XGBoost algorithm. Construct an objective function through its internal objective function construction unit, including a loss function and a regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model. Expand the loss function using Taylor's formula to obtain the final objective function, and perform secondary prediction based on the XGBoost algorithm to obtain the final prediction result.

[0044] Target energy supply information acquisition module 500: Determine the optimal energy supply strategy related to users according to the final prediction result, optimize energy supply, reduce waste, and improve energy utilization efficiency.

[0045] The prediction method includes the following steps: Step 1: Data acquisition and preprocessing Obtain the energy procurement cost, supply volume, and the energy load attributes of users (including temperature demand attributes and balance attributes) in the area where a wind farm is located during a specific time period.

[0046] Conduct an exploratory analysis on the obtained data. Perform a preliminary analysis on the dataset according to the time sequence and periodic law, and draw a scatter plot of feature variables to identify outliers. For the identified outliers, replace them with the previous value in the same column to ensure the continuity and periodicity of the data.

[0047] Step 2: Confirmation of key features Analyze the working principle and power input-output characteristics of the wind power generation system, and determine the key factors affecting wind power supply, such as wind speed, wind direction, etc.

[0048] Step 3: Extraction of feature variables Use mutual information entropy and principal component analysis (PCA) techniques to extract the key feature variables that significantly affect wind power supply from the original data. For example, average wind speed, maximum wind speed, etc.

[0049] Step 4: Initial prediction Construct a LightGBM model and use the extracted key features as input values for the initial prediction. Specifically, first sort according to the absolute value of the gradient of all samples, select the first a*100% of the samples as set A, and the remaining (1 - a)*100 samples as set AC. Randomly select b*100 samples as set B to generate a small gradient sample set, and perform an initial prediction based on the LightGBM model to obtain a preliminary prediction result.

[0050] Step 5: Secondary prediction Use the results of the initial prediction and the key features as input values, and apply the XGBoost algorithm for secondary prediction. Construct an objective function, including a loss function and a regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model. Expand the loss function using the Taylor formula to obtain the final objective function, and perform secondary prediction based on the XGBoost algorithm to obtain the final prediction result.

[0051] Step 6: Obtaining target energy supply information Determine the optimal energy supply strategy related to users according to the final prediction result, optimize energy supply, reduce waste, and improve energy utilization efficiency.

[0052] The parts not detailed in the present invention are prior arts. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, aiming to encompass all changes falling within the meaning and scope of equivalent elements within the present invention.

Claims

1. A renewable energy supply forecasting system in a power system, characterized by: include: A data acquisition and preprocessing module (100) is used to acquire data on energy procurement costs and supply volumes of renewable energy suppliers and energy load attributes of users at a target time; A feature extraction and selection module (200) is used to extract key features that significantly affect energy supply from the above data using mutual information entropy and principal component analysis (PCA) technology; The initial prediction module (300) is used to build a LightGBM model, perform initial prediction based on the above key features, and use the prediction results as new variables; A secondary prediction module (400) is used to use the result of the initial prediction and key features as input values, and use the XGBoost algorithm to perform secondary prediction to obtain a final prediction result; The target energy supply information obtaining module (500) is used to determine the optimal energy supply strategy related to the user according to the above-mentioned final prediction result.

2. The renewable energy supply prediction system in the power system according to claim 1, characterized in that: The feature extraction and selection module (200) has an abnormal value processing unit, which is used to identify, remove or replace abnormal values ​​in the original data.

3. The renewable energy supply prediction system in the power system according to claim 1, characterized in that: The initial prediction module (300) has a parameter updating unit for continuously adjusting and optimizing the parameters of the LightGBM model according to the prediction error.

4. The renewable energy supply prediction system in the power system according to claim 1, characterized in that: The secondary prediction module (400) has an objective function construction unit, which is used to construct an objective function including a loss function and a regularization term, and optimize the loss function by expanding the Taylor formula.

5. The renewable energy supply prediction system in the power system according to claim 1, characterized in that: The prediction method of the renewable energy supply prediction system in the power system comprises the following steps: S1: data acquisition step, used to obtain the energy procurement cost and supply volume of the renewable energy supplier at the target time and the energy load attributes of the user; S2: Data preprocessing step, which performs exploratory analysis on the acquired data and processes outliers to make the data continuous and periodic; S3: Key feature confirmation step, determine the key features according to the field of use; S4: feature variable extraction step, using mutual information entropy and principal component analysis (PCA) technology to extract key feature variables that significantly affect energy supply from the above data; S5: Initial prediction step, build the LightGBM model, use the extracted key features as input values ​​for initial prediction, and use the prediction results as new variables; S6: Secondary prediction step, taking the initial prediction result and key features as input values, using XGBoost algorithm for secondary prediction to obtain the final prediction result; S7: Target energy supply information acquisition step, determining the user-related optimal energy supply strategy based on the final prediction result.

6. The renewable energy supply prediction system in the power system according to claim 5, characterized in that: In S2, the power generation of renewable energy is considered as a continuous variable of one cycle, and the data set is exploratory analyzed according to the time sequence and periodicity law, and the scatter plot of the characteristic variables is drawn, and the abnormal values ​​and outliers are removed to ensure the continuity and periodicity of the data.

7. The renewable energy supply prediction system in the power system according to claim 5, characterized in that: In S4, information entropy uses numerical form to characterize the degree of uncertainty of the value of random variables and quantitatively describe the information content of random variables.

8. The renewable energy supply prediction system in the power system according to claim 5, characterized in that: In S5, GOSS sorts all samples according to their absolute gradient values, selects the first a*100% samples as set A, and the remaining (1-a)*100 samples as set AC; Randomly select b100 samples as set B, where the size of B is b|AC|, to generate a small gradient sample set; When calculating the information gain of a small gradient sample, a constant coefficient is multiplied, and the initial prediction is made based on the LightGBM model, and the prediction result is used as a new variable.

9. The renewable energy supply prediction system in the power system according to claim 5, characterized in that: In S6, the sample set xi (i=1,2,…,1974) is identified and a regression tree is constructed; Construct the objective function, including the loss function and the regularization term. The loss function is used to fit the training data, and the regularization term is used to simplify the model. The loss function is expanded using the Taylor formula to obtain the final objective function, and a secondary prediction is performed based on the XGBoost algorithm to obtain the final prediction result.

Images