New energy power grid extreme net load event prediction method and related device

Abstract

The invention discloses a prediction method for an extreme net load event of a new energy power grid and a related device. The prediction method comprises the following steps: acquiring historical net load data of the new energy power grid; carrying out per-unit processing on the historical net load data of the new energy power grid; performing p value analysis on the data subjected to per-unit processing, and judging whether the payload data in different years have correlation or not; subtracting a threshold value from the net loads of different years with correlation to obtain a difference value; and fitting the difference value through generalized Pareto distribution, and calculating the net load capacity once a hundred years. According to the method, fitting is carried out on the per-unit net load data through the generalized Pareto distribution, and the tail behavior of the load can be effectively captured, so that the extreme net load risk is accurately identified.

Description

Technical Field

[0001] This invention belongs to the field of new energy power grid technology, specifically relating to a method and related apparatus for predicting extreme net load events in new energy power grids. Background Technology

[0002] As the global energy structure shifts towards low-carbon and sustainable development, new energy sources such as wind and solar power are being integrated into the power grid on a large scale, and new energy power grids are gradually becoming an important part of the modern power system. This transformation is mainly due to the many advantages of new energy sources: environmental friendliness, reducing greenhouse gas emissions; renewable nature, ensuring resource sustainability; economic feasibility and technological advancements; and distributed generation and power system flexibility.

[0003] However, these advantages of new energy sources also bring new challenges. Due to the volatility of wind and solar power, new energy generation exhibits significant instability in actual power generation, and the net load (i.e., total load minus the portion supplied by new energy sources) also exhibits significant fluctuations. This places higher demands on grid load forecasting, real-time dispatching, and the management of extreme load events, especially when the penetration rate of new energy is high, making the grid more susceptible to the impact of extreme load events. Therefore, traditional load forecasting technologies can no longer fully meet the high-accuracy requirements of load forecasting in new energy grids, particularly when dealing with extreme load scenarios, where existing technologies have certain limitations.

[0004] In the context of renewable energy power grids, the importance of extreme load forecasting is increasingly prominent. However, current load forecasting technologies exhibit limitations when dealing with extreme load conditions in renewable energy power grids due to insufficient ability to capture extreme scenarios. Specifically, existing load forecasting methods are mainly based on traditional statistical models or machine learning algorithms, which have the following shortcomings when forecasting extreme loads in renewable energy power grids:

[0005] 1. Limitations of traditional statistical models: Commonly used time series models (such as ARIMA and LSTM) and basic load curve fitting methods are mainly based on the average trend of the data, usually assuming that the data follows a normal distribution, leading to insufficient prediction of extreme events. These models are difficult to reflect extreme load changes caused by fluctuations in new energy sources and cannot accurately assess the risks of extreme situations such as once-in-a-century events.

[0006] 2. Poor adaptability of machine learning models to extreme values: Although machine learning models (such as neural networks and random forests) have high accuracy in conventional load forecasting, their performance under extreme scenarios remains insufficient. Machine learning models typically rely on large-scale historical data for training, while extreme scenarios in renewable energy power grids are relatively rare, making it difficult to provide sufficient extreme data samples to improve prediction accuracy. Consequently, it is difficult to capture the fluctuation characteristics of renewable energy power grids under extreme load conditions.

[0007] 3. Lack of extreme load modeling: Existing technologies rarely model extreme loads, especially in renewable energy grids, where the distribution characteristics of extreme loads have not been fully utilized or explored. Although extreme load scenarios (such as a sudden and significant drop in renewable energy output or a sudden increase in load) occur infrequently, they have a significant impact on grid security once they occur. Therefore, specialized extreme value modeling methods are needed for reasonable risk assessment and early warning.

[0008] 4. Lack of effective strategies to address the uncertainties of renewable energy sources: The net load characteristics of renewable energy power grids are significantly affected by meteorological factors. This uncertainty makes it difficult for traditional methods to cope with renewable energy fluctuations under extreme weather conditions and to accurately predict extreme load values. Existing forecasting methods lack comprehensive modeling of extreme weather and other climate variables, making it impossible to accurately forecast loads when renewable energy output fluctuates significantly.

[0009] The accuracy and reliability of load forecasting for renewable energy power grids are crucial, especially under extreme scenarios, as they directly impact grid security and operational stability. Therefore, to address extreme load conditions caused by the volatility of renewable energy sources, it is necessary to develop a new method for forecasting extreme net load events in the power grid. Summary of the Invention

[0010] To overcome the technical problem of unreasonable load forecasting in existing renewable energy power grids, the present invention aims to provide a method and related apparatus for predicting extreme net load events in renewable energy power grids. This method is based on the generalized Pareto distribution and, by fitting the extreme distribution characteristics of the net load curve, can achieve reasonable prediction of extreme scenarios such as once-in-a-century events in the power grid, providing more efficient decision support for the power grid and improving the risk prevention and control capabilities and stability of renewable energy power grids.

[0011] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0012] The first aspect of this invention provides a method for predicting extreme net load events in a new energy power grid, comprising the following steps:

[0013] Obtain historical net load data for the new energy power grid;

[0014] Standardize the historical net load data of the new energy power grid;

[0015] A p-value analysis was performed on the normalized data to determine whether the net load data of different years were correlated.

[0016] The difference is obtained by subtracting the threshold from the net load of different correlated years;

[0017] By fitting the difference to a generalized Pareto distribution, the net load for a once-in-a-century event was calculated.

[0018] Furthermore, the historical net load data of the new energy power grid includes the hourly load of the grid and the amount of renewable energy generated within a set time period, covering multiple years.

[0019] Furthermore, the historical net load data of the new energy power grid is normalized using the following formula:

[0020]

[0021] Where X represents historical net load data, μ represents the mean of the data, and σ represents the standard deviation.

[0022] Furthermore, p-value analysis was performed on the standardized data to determine whether the net load data for different years were correlated, including the following steps:

[0023] The p-value was calculated using statistical testing methods on the standardized historical net load data. If the p-value was less than or equal to the set value, the net load data of different years were considered to be correlated; the set value was 0.05.

[0024] Furthermore, a threshold is determined based on the average excess quantity function. After the threshold is satisfied, the average excess quantity function exhibits a linear trend.

[0025] Furthermore, the average excess quantity function e(u) is:

[0026]

[0027] Where X is the net load, u is the threshold, and N is the threshold value. u The number of samples exceeding the threshold u is represented by ξ, which is a shape parameter controlling the thickness of the tail, and σ is a scale parameter. i Let be the net load for year i, where i is the length of time in years.

[0028] Furthermore, the generalized Pareto distribution takes the form of:

[0029]

[0030] In the formula, ξ is a shape parameter that controls the thickness of the tail; σ is a scale parameter; and x is a variable.

[0031] A first aspect of the present invention provides a prediction system for extreme net load events in a renewable energy power grid, comprising:

[0032] The data acquisition module is used to acquire historical net load data of the new energy power grid;

[0033] The per-unit processing module is used to standardize historical net load data of the new energy power grid.

[0034] The correlation judgment module is used to perform p-value analysis on the standardized data to determine whether the net load data of different years are correlated.

[0035] The difference calculation module is used to subtract a threshold from the net load between correlated years to obtain the difference.

[0036] The fitting module is used to fit the difference using a generalized Pareto distribution to calculate the net load that occurs once every 100 years.

[0037] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting extreme net load events in the new energy power grid.

[0038] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for predicting extreme net load events in the new energy power grid.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. Accurate Identification of Extreme Net Load Risks: This invention fits the per-unit net load data using a generalized Pareto distribution, effectively capturing the tail behavior of the load and thus accurately identifying extreme net load risks. Through this method, grid operators can better understand the frequency and intensity of extreme load events, providing a basis for developing appropriate contingency plans. This overcomes the lack of systematic analysis and prediction of extreme net load conditions in power systems in existing technologies, especially in the context of large fluctuations in renewable energy.

[0041] 2. Quantifying the correlation of net load between different years and providing consistency verification: The p-value analysis of this invention is not only used to verify the correlation of net load between different years, but also to analyze data consistency, thereby identifying whether extreme events show long-term trends due to factors such as climate change. Through this improvement, the power grid can more flexibly adjust load forecasting methods to adapt to changing climate conditions and their impact on renewable energy output, thereby enhancing the stability of the power system.

[0042] 3. Providing scientific forecasts of once-in-a-century extreme loads for rational power grid planning: Based on the generalized Pareto distribution, this invention can quantitatively predict once-in-a-century extreme net load events. Power system planning, especially in capacity design and risk management, relies heavily on accurate estimates of extreme loads. The once-in-a-century extreme load forecasts provided by this invention offer data support for rational capacity reserve design, emergency resource allocation, and power grid planning, reducing the risk of power outages caused by overload and overcoming the problem that existing technologies often use historical averages for design, failing to fully consider the impact of extreme situations.

[0043] The various technical steps of this invention work together to make effective improvements in all aspects, from data preprocessing, distribution fitting, threshold selection, extreme value analysis to extreme load prediction. It is scientific, universal, and operable, greatly enhancing the power system's ability to cope with extreme net load events and providing a solid theoretical and practical foundation for the future development of renewable energy grids.

[0044] Furthermore, this invention employs threshold selection for net load data and tail estimation of the average excess function, particularly utilizing a generalized Pareto distribution to conduct detailed analysis of extreme data exceeding the threshold. This method ensures the accuracy of predictions under highly fluctuating load events, enabling the power grid to effectively prevent and respond to the impact of extreme weather or load conditions on grid stability. It overcomes the limitations of existing technologies that often use only a single distribution to model all load data, failing to adapt to the specificity of extreme events. Attached Figure Description

[0045] Figure 1 This is a flowchart of the method for predicting extreme net load events in the new energy power grid according to the present invention;

[0046] Figure 2 This is a schematic diagram of the prediction system for extreme net load events in the new energy power grid according to the present invention. Detailed Implementation

[0047] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0048] In power systems where the proportion of renewable energy is constantly increasing, accurate prediction of extreme net load values ​​is crucial. Net load refers to the total system load minus the generation of renewable energy sources (such as wind and solar power). The output of these renewable energy sources is often affected by weather conditions, exhibiting significant volatility. Especially under extreme weather or high load demand conditions, changes in net load can lead to grid instability. Therefore, this invention employs the Generalized Pareto Distribution (GPD) method to model and predict the extreme net load of the power grid, providing a reference for power dispatching and decision-making.

[0049] See Figure 1 The present invention provides a method for predicting extreme net load events in a new energy power grid, comprising the following steps:

[0050] The first step is to collect historical net load data. This data typically includes the hourly load of the grid and renewable energy generation over a specific time period. The collected data should cover multiple years to ensure the representativeness and validity of the analysis.

[0051] The second step is to standardize the historical net load data.

[0052] Per-unit normalization converts historical net load data into a dimensionless form for easier comparison and analysis. Specifically, the following formula is used for per-unit normalization:

[0053]

[0054] Where X represents historical net load data, μ is the mean of the data, and σ is the standard deviation. After normalization, the mean of the data is 0, and the standard deviation is 1. This step helps to eliminate seasonality or biases between different load levels, as well as the impact of differences in data magnitude on subsequent analysis, making the data more comparable.

[0055] The third step, after standardization, is to perform p-value analysis on the data to assess the correlation between net loads across different years. P-value analysis is a statistical method used to test whether a hypothesis holds true. In this context, it primarily involves hypothesis testing to determine whether there are significant differences in net load data across different years. The steps for hypothesis testing are as follows:

[0056] 1. Calculate the p-value: Use statistical testing methods (such as t-test or ANOVA) to calculate the p-value for the standardized historical net load data. The p-value represents the probability of observing the current or more extreme outcome, assuming the null hypothesis is true.

[0057] 2. Set the significance level: Usually, a significance level (alpha) of 0.05 is chosen. When the p-value is less than or equal to 0.05, the null hypothesis is rejected, and it is considered that there is a significant difference between the net load data of different years, that is, the net loads of different years are not correlated; when the p-value is greater than 0.05, the null hypothesis is accepted, and it is considered that there is no significant difference, and the net loads of different years are correlated.

[0058] For example, this analysis of the Northwest China power grid concluded that net load data showed a significant correlation in years other than 2020. This finding is significant because it indicates that the load patterns of the power system are stable across different years, potentially influenced by similar climatic conditions and electricity consumption habits. This provides a basis for subsequent extreme value analysis, demonstrating that relatively accurate predictions can be made based on historical data.

[0059] After completing the p-value analysis, threshold selection and parameter estimation for values ​​exceeding the threshold are performed to identify extreme load events. Threshold selection is a crucial step in extreme value analysis, typically determined using empirical methods or statistical tests (such as the Peirce method). To ensure model accuracy, this invention employs the average excess quantity function e(u) for threshold determination, defined as:

[0060]

[0061] Where X is the net load, u is the threshold, and N is the threshold value. u The number of samples exceeding the threshold u is represented by ξ, which is a shape parameter controlling the thickness of the tail, and σ is a scale parameter. i Let be the net load for year i, where i is the length of time in years.

[0062] From the above formula, we can see that e(u) should be a linear function of u. Therefore, by using different thresholds, if e(u) shows a linear trend after a certain threshold u0, then u0 is selected as a suitable threshold. Determining a suitable threshold helps to capture the extreme characteristics of the data and avoids treating normal fluctuations as extreme events.

[0063] After selecting a threshold, the difference is obtained by subtracting the threshold from the net load of correlated years. This difference is then fitted using a generalized Pareto distribution (GPD) to calculate the once-in-a-century net load. The GPD is used to describe the tail behavior of extreme values, and its form is as follows:

[0064]

[0065] In the formula, ξ is a shape parameter that controls the thickness of the tail; σ is a scale parameter that controls the scale or width of the distribution and determines the rate of change of the tail; and x is a variable.

[0066] To calculate the once-in-a-century extreme net load, the distribution characteristics of events exceeding a threshold are defined. Therefore, the concept of recurrence level is introduced to quantify whether a particular renewable energy output event can be defined as a "once-in-a-century" event. Specifically, assuming a time length of T years, the recurrence level u(T) is defined to satisfy the following condition: within the time range of T years, the average number of events exceeding the set threshold u(T) is 1, formally expressed as:

[0067]

[0068] Wherein, I{X i >u(T)} is an indicator function, where i is the number of years in time, indicating when X i The value of u(T) is 1 when u(T) is greater than or equal to 0 otherwise. The recurrence level u(T) can be viewed as the expected number of times an event exceeds a threshold u(T) within a time period T. In this way, the frequency of extreme events within a specific time period can be quantified.

[0069] The above methods not only accurately identify and predict extreme net load conditions but also ensure the stability and security of the power system. This technical solution has significant practical application value and research significance in the context of renewable energy gradually becoming mainstream. Through scientific statistical analysis and modeling techniques, it provides strong support for the future development of the power system, ensuring effective response to extreme climate change.

[0070] This invention utilizes the Generalized Pareto Distribution (GEV) to model the net load generation function curve, identifying and predicting extreme load conditions. Per-unit processing is employed to standardize historical net load data, eliminating seasonality or biases between different load levels, as well as the impact of data size differences on the analysis results. A p-value analysis method is introduced to assess the correlation of net load between different years, ensuring the accuracy of model predictions and making the data highly comparable. By determining reasonable load thresholds, extreme load events are identified, and the shape, location, and scale parameters of the GEV are determined using maximum likelihood estimation to ensure the accuracy of extreme event modeling. Furthermore, the GEV is used to fit the processed data, identifying and analyzing the long-tail characteristics of the system net load, calculating once-in-a-century extreme load volumes, proposing the concept of recurrence level, and quantifying the frequency of extreme load events within a specific time period (within a given year period), providing decision support for the long-term stability of power system dispatch.

[0071] As the proportion of renewable energy in the power system gradually increases, the system's ability to withstand fluctuations has become a major challenge to ensuring system stability. Especially in the context of frequent extreme weather events, the system needs to be able to predict and respond to extreme net load events. Traditional forecasting methods, neglecting the characteristics of extreme values, are insufficient to meet current dispatching needs. Furthermore, existing net load forecasting methods are mostly based on historical statistical averages or use manually set simple limits as early warning standards, which ignore the long-tail distribution characteristics of extreme loads, making it difficult to accurately capture and predict extreme events, especially in regions with high renewable energy volatility. In addition, many methods are only applicable to short-term load fluctuation forecasting and cannot provide a reference for long-term extreme loads. Therefore, this invention employs extreme value theory to model and predict the extreme net load of the power grid to support future dispatching decisions.

[0072] Example 1

[0073] Table 1 Net Load of Northwest Power Grid

[0074]

[0075]

[0076] The actual maximum net load for the six years from 2018 to 2023 was 295,194 MW, while the predicted value was 294,710 MW, with a prediction error of 0.164%.

[0077] This invention first verifies the correlation between historical data through statistical analysis, then models the long-tailed distribution characteristics of net load using extreme value theory, and further determines the probability distribution of extreme events. Each step of the reasoning process is based on a mathematical model, ensuring the scientific rigor and soundness of the conclusions.

[0078] See Figure 2 Another embodiment of the present invention provides a prediction system for extreme net load events in a new energy power grid, comprising:

[0079] Another embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: when the processor executes the computer program, it implements the method for predicting extreme net load events of the new energy power grid.

[0080] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for predicting extreme net load events in the new energy power grid.

[0081] This invention is primarily applicable to power systems with a high proportion of renewable energy integration, and is suitable for scenarios involving extreme load forecasting in dispatching and planning. Compared to traditional methods, this approach can significantly improve the accuracy of extreme net load forecasting and helps to deploy load balancing measures in advance to reduce grid fluctuation risks.

[0082] The above description is only of the preferred embodiment of the present invention and should not be construed as limiting the scope of the claims. The present invention is not limited to the above embodiments, and variations in its specific structure are permitted. All variations made within the scope of the independent claims of the present invention are also within the scope of protection of the present invention.

[0083] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

Claims

1. A method for predicting extreme net load events of a new energy power grid, characterized in that: The following steps are involved: Obtain historical net load data of new energy power grid; Standardize the historical net load data of the new energy power grid; Perform p-value analysis on the normalized data to determine whether the net load data of different years are correlated; Subtract the threshold from the net load between different years of correlation to obtain the difference; The difference is fitted by generalized Pareto distribution and the net load capacity of once in a hundred years is calculated.

2. The method for predicting extreme net load events of a new energy power grid according to claim 1, characterized in that: The historical net load data of the new energy grid includes the hourly load and renewable energy power generation of the grid within a set time period, and covers multiple years.

3. The method for predicting extreme net load events of a new energy power grid according to claim 1, characterized in that: The historical net load data of the new energy power grid is normalized using the following formula: Among them, X is the historical net load data, μ is the mean of the data, and σ is the standard deviation.

4. The method for predicting extreme net load events of a new energy power grid according to claim 1, characterized in that: Perform p-value analysis on the normalized data to determine whether the net load data of different years are correlated, including the following steps: The p-value is calculated for the normalized historical net load data using a statistical test method. If the p-value is less than or equal to a set value, the net load data of different years are considered to be correlated; the set value is 0.

05.

5. The method for predicting extreme net load events of a new energy power grid according to claim 1, characterized in that: A threshold is determined according to the average excess function, and the threshold is met. After the threshold, the average excess function presents a linear trend.

6. The method for predicting extreme net load events of a new energy power grid according to claim 5, characterized in that: The average excess function e(u) is: Where X is the net load, u is the threshold, N u represents the number of samples exceeding the threshold u, ξ is the shape parameter controlling the thickness of the tail, σ is the scale parameter, X i is the net load in the i-th year, i is the length of time in years.

7. The method for predicting extreme net load events of a new energy power grid according to claim 1, characterized in that: The generalized Pareto distribution is: Where ξ is the shape parameter that controls the thickness of the tail; σ is the scale parameter, and x is the variable.

8. A prediction system for extreme net load events of a new energy power grid, characterized in that: include: A data acquisition module is used to obtain historical net load data of the new energy power grid; A per-unit processing module is used to perform per-unit processing on the historical net load data of the new energy power grid; The correlation judgment module is used to perform p-value analysis on the normalized data to determine whether the net load data of different years are correlated; A difference calculation module is used to subtract a threshold from the net loads between different years with correlation to obtain a difference; The fitting module is used to fit the difference value through the generalized Pareto distribution and calculate the net load of once in a hundred years.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for predicting extreme net load events of a new energy power grid as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for predicting extreme net load events of a new energy power grid as described in any one of claims 1 to 7 is implemented.

Images