Optimized combination method of new energy power prediction model

Abstract

The invention provides an optimal combination method of a new energy power prediction model, and relates to the technical field of new energy power prediction. Obtaining characteristic parameters of the station; evaluating the characteristic parameters to obtain a characteristic evaluation result, and taking the characteristic evaluation result as the station characteristics of the target station; aiming at a plurality of pre-stored power prediction models, carrying out weight distribution on each pre-stored power prediction model by utilizing the station characteristics; selecting the most suitable model or model combination by using a model selection function according to the station characteristics and the model weight; for the obtained model or model combination, predicting an error of the model or model combination; according to an error evaluation result, adjusting model parameters to obtain an optimized power prediction model; the prediction model can be selected according to the actual conditions of different stations, and the pertinence of power prediction is high.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy power prediction, and in particular to an optimization combination method of a new energy power prediction model. Background Art

[0002] With the continuous increase in the installed capacity of renewable energy power generation, the proportion of renewable energy output has increased year by year. In recent years, with the large-scale access of renewable energy and the development of power trading, there is a large random uncertainty in the transmission lines and load sides. At the same time, it is affected by factors such as wind speed, light intensity and meteorology, which makes it difficult to accurately know the real-time operation status of electricity in advance, thus bringing great challenges to the operation of the power system. At the same time, wind power and solar power generation have the characteristics of natural distribution. They are mainly distributed in vast and sparsely populated areas such as Northwest, North China and Northeast China, and are greatly affected by the seasonality of resources. In addition, due to the relatively late development of wind and solar energy resources in my country, many places have not yet fully formed mature business and technical models to support the construction and development of large-scale wind farms. These factors increase the uncertainty of renewable energy power generation, making the power supply of the power grid more complicated and unpredictable, and the prediction of renewable energy power generation is crucial.

[0003] At present, the power prediction methods for renewable energy power generation can be divided into a variety of algorithm models, including short- and medium-term deterministic predictions, short-term random predictions, and medium- and long-term probabilistic predictions. However, the existing technology cannot recommend the optimal combination of prediction schemes and algorithm models for different sites, which limits the targeted nature of power predictions. Summary of the invention

[0004] The purpose of the present invention is to provide an optimization combination method for a new energy power prediction model, which can select a prediction model according to the actual situation of different sites, and the power prediction is highly targeted.

[0005] The embodiments of the present invention are implemented by the following technical solutions:

[0006] An optimization combination method for a new energy power prediction model, comprising:

[0007] Obtain characteristic parameters of the station;

[0008] Evaluate the characteristic parameters to obtain characteristic evaluation results, and use the characteristic evaluation results as the station characteristics of the target station;

[0009] For multiple pre-stored power prediction models, the site characteristics are used to assign weights to each pre-stored power prediction model;

[0010] Use the model selection function to select the most appropriate model or model combination based on site characteristics and model weights;

[0011] For the obtained model or model combination, predict the error of the model or model combination;

[0012] According to the error evaluation results, the model parameters are adjusted to obtain the optimized power prediction model.

[0013] Furthermore, the characteristic parameters of the station include regional, seasonal, environmental and climate characteristics, and are represented by a set of characteristic vectors.

[0014] Furthermore, the characteristic vector is expressed by the formula:

[0015] FeatureVector=[F region ,F season ,F environment ,F clintate ].

[0016] Furthermore, the evaluation step includes:

[0017] Collect the original data of the target station within a certain historical period;

[0018] Extract target features from raw data;

[0019] For each target feature, a statistical method is used to evaluate the relationship between the target feature and the power generation to obtain an evaluation result of the feature parameter;

[0020] The several evaluation results with the highest correlation are taken as the site characteristics of the target electric field.

[0021] Furthermore, the statistical method is specifically a correlation coefficient method, and the specific calculation formula of the method is:

[0022]

[0023] Among them, r is the Pearson correlation coefficient, X and Y are two variables, and is the mean of the two variables and n is the number of data points.

[0024] Furthermore, the statistical method is specifically a regression analysis method, and the specific calculation formula of the method is:

[0025] Y=β0+β1X1+β2X2+···+β n X n +ε;

[0026] Among them, Y is the power generation, X1~X n is the target feature, β1~β n is the regression coefficient, and ε is the error. Further, the calculation formula of the weight distribution is:

[0027] Wi =f(I i ,M i )=a·I i +b·M i

[0028] Among them, W i is the weight of the ith model, I i is the matching degree between the ith model and the station characteristics, M i is the historical performance or feature importance evaluation result of the ith model, f(·) is the weight assignment function, and a and b are constants used to adjust the relative influence of the matching degree and feature importance.

[0029] Furthermore, the calculation formula of the model selection function is:

[0030]

[0031] Where n is the number of target models, W i is the weight of the i-th model, Model i (InputData) is the i-th model.

[0032] Furthermore, the error prediction formula of the model or model combination is:

[0033]

[0034] Where N is the number of data points, PredictedPower t is the predicted power at the tth time point, ActualPower t is the actual power at the tth time point.

[0035] Furthermore, the optimized power prediction model is pushed to the user's smart terminal.

[0036] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:

[0037] The present invention can recommend prediction models for different sites, thereby improving the pertinence of power prediction and minimizing the cost risk of power generation enterprises;

[0038] In addition, model parameters are continuously adjusted and optimized based on the model prediction error to ensure that the model maintains high prediction performance over time. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0040] Figure 1 A schematic diagram of a method flow chart provided in Example 1 of the present invention;

[0041] Figure 2 This is a schematic diagram of the method flow provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0043] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0045] In the description of the present invention, it should be noted that if the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" and the like appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the accompanying drawings, or is the orientation or position relationship in which the product of the application is usually placed when used. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0046] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0047] Example 1

[0048] As attached Figure 1 An optimization combination method of a new energy power prediction model shown includes:

[0049] Obtain characteristic parameters of the station;

[0050] In particular, the characteristic parameters of the station include regional, seasonal, environmental and climatic characteristics, and are represented by a set of characteristic vectors; and are expressed by the formula:

[0051] FeatureVector=[F region ,F season ,F environment ,F clintate ].

[0052] That is, FeatureVector is the feature vector, F regoon is the regional characteristic parameter, F season is the seasonal characteristic parameter, F environment is the environmental characteristic parameter, F clintate is the climate characteristic parameter;

[0053] Since feature vectors allow complex feature sets to be simplified into a set of numerical values, this facilitates subsequent analysis and calculations, and in this way, large amounts of data can be more easily managed and manipulated, especially when a large number of features are involved in this method; in addition, by vectorizing feature parameters, numerical comparisons and operations can be performed, thereby supporting quantitative analysis; and for subsequent machine learning algorithms, input data usually needs to be standardized, so feature vectorization can make different types of input features have the same dimensions and format, which is convenient for unified processing and input into the model;

[0054] In addition, for predictive models, feature vectors provide a structured way to organize input data, allowing the model to efficiently learn the mapping relationship between input and output; and after feature vectorization, the efficiency of matrix operations can be used to perform fast calculations, greatly improving the calculation speed.

[0055] Evaluate the characteristic parameters to obtain characteristic evaluation results, and use the characteristic evaluation results as the station characteristics of the target station;

[0056] By evaluating the characteristic parameters, we can screen out those characteristics that are highly correlated with power generation, so as to build a more accurate prediction model, which helps to improve the prediction performance of the model and reduce errors. In addition, during the evaluation process, not only the direct relationship between the characteristics and power generation is taken into account, but also the importance and historical performance of the characteristics. This enables the model to be optimized according to the degree of influence of each characteristic on power generation prediction, so as to select the model or model combination that best suits the current site characteristics.

[0057] And because each new energy station has its own unique geographical, climatic, environmental and other characteristics, evaluating these characteristic parameters can help identify the uniqueness of each station. Based on this information, the system can recommend the most suitable prediction model for each station to achieve personalized prediction services. In the process of evaluating characteristic parameters, the most representative and influential features will be selected, which reduces the amount of data that needs to be processed and reduces the computational complexity, thereby saving computing resources and time costs.

[0058] For multiple pre-stored power prediction models, the site characteristics are used to assign weights to each pre-stored power prediction model;

[0059] The pre-stored power prediction models include short- and medium-term deterministic prediction, short-term random prediction, medium- and long-term probabilistic prediction models, neural network models based on deep learning, and integrated learning methods;

[0060] In particular, the calculation formula for the weight distribution is:

[0061] W i =f(I i ,M i )=a·I i +b·M i

[0062] Among them, W i is the weight of the ith model, I i is the matching degree between the ith model and the station characteristics, M iis the historical performance or feature importance evaluation result of the ith model, f(·) is the weight assignment function, a and b are constants used to adjust the relative influence of the matching degree and feature importance; the purpose of this weight assignment process is to determine which models are more suitable for the current situation based on the characteristics of the site. Through such calculations, a weight value can be assigned to each pre-stored power prediction model, which reflects the adaptability of the model to the characteristics of the current site; then, using these weight values, combined with the site characteristic evaluation, the model or model combination that best suits the current site is selected through the model selection function. This weight-based model selection method can recommend the most appropriate prediction model based on the different characteristics of each site, thereby improving the pertinence of power prediction.

[0063] In addition, the weights can be recalculated based on the performance of the model in the recent period, which can make the prediction model more adaptable to changing environmental conditions;

[0064] The method of dynamically adjusting weights means recalculating the weights based on the performance of the model in the recent period. Specifically, this means that the system will regularly evaluate the prediction errors of each model, and then adjust the model weights based on these errors. The basic idea of ​​weight adjustment is: if a model has performed well in the past period of time, the weight of the model may increase; conversely, if the model performs poorly, its weight may decrease.

[0065] This method enables the prediction capability of the entire prediction system to self-adjust with changes in time and environment to achieve the best prediction effect. In actual applications, weight adjustment can be achieved in a variety of ways, such as using online learning algorithms, sliding window methods and other technologies to continuously update model weights. The advantage of this is that the system can respond to the latest data changes in a timely manner and improve the accuracy and timeliness of the prediction.

[0066] Use the model selection function to select the most appropriate model or model combination based on site characteristics and model weights;

[0067] In particular, the calculation formula of the model selection function is:

[0068]

[0069] Where n is the number of target models, W i is the weight of the i-th model, Model i (InputData) is the i-th model;

[0070] It should be noted that if a site exhibits different characteristics in different time periods, a single model may not be able to capture all the characteristics well, and a combination of multiple models can better cope with this situation. In this case, the selection function will select those models with higher weights and that best match the characteristics of the current site for combination.

[0071] For the obtained model or model combination, predict the error of the model or model combination;

[0072] The prediction error can be used to evaluate the performance of a model or model combination at a specific site. The error value can be used as an indicator to measure the accuracy of the prediction, helping to determine whether the current model combination is suitable for the characteristics of the current site. In addition, the calculated error can be used as a feedback signal to adjust model parameters or select different model combinations, that is, the prediction error is not only the basis for model selection, but also the basis for continuous improvement of the model. By periodically evaluating the error, it is possible to promptly detect the decline in model performance and take appropriate measures to adjust it to ensure the accuracy of the prediction. It should be noted that error evaluation provides a method for monitoring model performance. When the performance of the model deviates from expectations, the error value will increase, which indicates that the model needs to be checked or updated.

[0073] In particular, the error prediction formula of the model or model combination is:

[0074]

[0075] Where N is the number of data points, PredictedPower t is the predicted power at the tth time point, ActualPower t is the actual power at the tth time point.

[0076] According to the error evaluation results, the model parameters are adjusted to obtain the optimized power prediction model;

[0077] The optimized power prediction model is pushed to the user's smart terminal. Pushing it to the smart terminal allows the user to obtain the latest prediction results in real time, which is particularly important for scenarios that require immediate response, such as power dispatching and grid balance. In addition, the user can directly use the latest optimized model on their device without manually downloading or updating the model, which simplifies the operation process and improves the user experience. Moreover, each site has different characteristics, and the service pushed to the smart terminal can be customized according to the characteristics of the specific site, providing users with personalized power prediction services.

[0078] In addition, through smart terminals, users can remotely monitor the status of power plants and make corresponding management decisions based on the prediction results, enhancing the ability of remote operation; applications on smart terminals can display prediction results and other relevant information to help users make more informed decisions and improve operational efficiency; smart terminals usually have good Internet connections, which makes it easy for them to integrate with other systems, such as energy management systems, trading platforms, etc., to form a complete solution; and data pushed to smart terminals can be backed up in the cloud to ensure data security, while also facilitating synchronization between different devices.

[0079] It should be noted that pushing new versions of models or combinations to smart terminals can ensure that users always use the latest technical support and enjoy the performance improvements and new features brought by software updates in a timely manner.

[0080] Example 2

[0081] According to the attached Figure 2 As shown, the evaluation step includes:

[0082] Collect the original data of the target station within a certain historical period; by collecting the historical power generation data, weather forecast data, environmental data and climate data of the target station, a comprehensive original data set can be obtained, which helps to ensure that the model can be trained based on sufficient information, thereby improving the accuracy of the prediction;

[0083] Extract target features from raw data: Extract target features that have a direct impact on power generation from raw data, such as average wind speed, solar radiation intensity, etc. This process helps remove irrelevant data and focus on variables that are critical to the prediction results, thereby improving the efficiency and accuracy of the model;

[0084] For each target feature, a statistical method is used to evaluate the relationship between the target feature and the power generation to obtain the evaluation results of the feature parameters; the features with the highest correlation with the power generation can be identified to provide a basis for model selection, among which statistical evaluation helps to screen out the most influential features and avoid overfitting;

[0085] The several evaluation results with the highest correlation are taken as the site characteristics of the target electric field.

[0086] In addition, the statistical method used is the correlation coefficient or regression analysis method, where the specific calculation formula of the correlation coefficient method is:

[0087]

[0088] Among them, r is the Pearson correlation coefficient, X and Y are two variables, which can be any two sets of data. For example, in the context of renewable energy power prediction, X may represent environmental characteristic parameters such as average wind speed, solar radiation intensity or temperature change, while Y represents power generation; and is the average of the two variables, n is the number of data points;

[0089] Moreover, the Pearson correlation coefficient is a value between -1 and 1, which is used to measure the strength of the linear relationship between two variables. It is simple to calculate, intuitive to explain, easy to understand and apply. In addition, by calculating the correlation coefficient, the strength and direction of the relationship between the two variables can be directly obtained. A positive value indicates a positive correlation, and a negative value indicates a negative correlation. The larger the absolute value, the stronger the relationship. The size of the correlation coefficient is independent of the measurement scale of the variable, which makes it still valid when comparing data at different scales.

[0090] In addition, the correlation coefficient can be used as a means of preliminary screening of features. By calculating the correlation between the features and the target variable, the features with high correlation with the target variable can be quickly found.

[0091] The specific calculation formula of the regression analysis method is:

[0092] Y=β0+β1X1+β2X2+···+β n X n +ε;

[0093] Among them, Y is the power generation, X1~X n is the target feature, β1~β n is the regression coefficient, ε is the error;

[0094] Regression analysis can not only reveal the relationship between independent variables (features) and dependent variables (power generation), but also establish a mathematical model for predicting unknown data. The regression coefficient (β1-βn) obtained through regression analysis can explain the contribution of each independent variable to the dependent variable and help understand which features have the greatest impact on power generation.

[0095] In addition, regression analysis supports the simultaneous consideration of the impact of multiple independent variables on the dependent variable, which is more common in the real world because many phenomena are the result of the combined effect of multiple factors. Regression analysis provides a variety of diagnostic tools, such as residual analysis, collinearity diagnosis, etc., which can help evaluate the fit of the model, identify potential problems and correct them. The regression model can be flexibly expanded according to actual conditions, such as adding interaction terms, nonlinear terms, etc., to make it more in line with the distribution law of actual data.

[0096] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An optimization combination method for a new energy power prediction model, characterized in that: include: Obtain characteristic parameters of the station; These characteristic parameters are evaluated to obtain characteristic evaluation results, and the characteristic evaluation results are used as the station characteristics of the target station; For multiple pre-stored power prediction models, the site characteristics are used to assign weights to each pre-stored power prediction model; Use the model selection function to select the most appropriate model or model combination based on site characteristics and model weights; For the obtained model or model combination, predict the error of the model or model combination; According to the error evaluation results, the model parameters are adjusted to obtain the optimized power prediction model.

2. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The characteristic parameters of the station include regional, seasonal, environmental and climate characteristics, and are represented by a set of characteristic vectors.

3. The optimization combination method of the new energy power prediction model according to claim 2, characterized in that: The eigenvector is expressed by the formula: FeatureVector=[F region ,F season ,F environment ,F clintate ]。 4. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The evaluation steps include: Collect the original data of the target station within a certain historical period; Extract target features from raw data; For each target feature, a statistical method is used to evaluate the relationship between the target feature and the power generation to obtain an evaluation result of the feature parameter; The several evaluation results with the highest correlation are taken as the site characteristics of the target electric field.

5. The optimization combination method of the new energy power prediction model according to claim 4, characterized in that: The statistical method is specifically a correlation coefficient method, and the specific calculation formula of the method is: Among them, r is the Pearson correlation coefficient, X and Y are two variables, and is the mean of the two variables and n is the number of data points.

6. The optimization combination method of the new energy power prediction model according to claim 4, characterized in that: The statistical method is specifically a regression analysis method, and the specific calculation formula of the method is: Y=β0+β1X1+β2X2+···+β n X n +e; Among them, Y is the power generation, X1~X n is the target feature, β1~β n is the regression coefficient and ε is the error.

7. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The calculation formula for the weight distribution is: W i =f(I i ,M i )=a*I i +b*M i Among them, W i is the weight of the ith model, I i is the matching degree between the ith model and the station characteristics, M i is the historical performance or feature importance evaluation result of the ith model, f(·) is the weight assignment function, and a and b are constants used to adjust the relative influence of the matching degree and feature importance.

8. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The calculation formula of the model selection function is: Where n is the number of target models, W i is the weight of the i-th model, Model i (InputData) is the i-th model.

9. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The error prediction formula of the model or model combination is: Where N is the number of data points, PredictedPower t is the predicted power at the tth time point, ActualPower t is the actual power at the tth time point.

10. The optimization combination method of the new energy power prediction model according to claim 1, characterized in that: The optimized power prediction model is pushed to the user's smart terminal.

Images