New energy access load prediction precision improvement method, system and device

Abstract

The invention discloses a load prediction precision improvement method considering new energy access. The method comprises the following steps: establishing a sample data set; training a whole-network load direct prediction model by using the data of the sample data set to obtain a whole-network load direct prediction result A; combining the data of the sample data set, the plant power consumption rate data of the thermal power plant and the plant power consumption rate data of the new energy plant to train a power consumption load prediction model to obtain a whole-network load indirect prediction result B; and integrating the whole-network load direct prediction result A and the whole-network load indirect prediction result B to obtain an optimal prediction result. According to the invention, for the problem that the wide-area distributed new energy access aggravates the system load fluctuation and prediction difficulty, the processing of the station service power change caused by the distributed new energy access is added in the system load prediction, and the system load prediction precision in the wide-area distributed new energy access scene is improved. And accurate and reliable boundary data are provided for scheduling operation and market optimization.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system load forecasting, and specifically relates to a method, system and device for improving load forecasting accuracy considering the access of new energy sources. Background Art

[0002] The penetration rate of distributed renewable energy continues to increase. Although renewable energy power generation has no impact on the electricity demand of the whole society, the system load used for actual business such as dispatch planning and power market in the power system is all machine-side power generation rather than grid-connected power, that is, it includes power consumption of power plants. The power consumption rate of renewable energy power generation is much lower than that of traditional thermal power. The fluctuation of renewable energy power generation will cause a large change in power consumption, thereby exacerbating the fluctuation of system load and further increasing the difficulty of system load forecasting. The system load forecast result, as the key boundary data of the power market, is an important guarantee for the safe, stable and economic operation of the power market.

[0003] In traditional situations, renewable energy generation accounts for a small proportion of all power generation, and the system load is mainly composed of traditional thermal power. The power consumption of the plant is approximately proportional to the system load, which will not affect the regularity of the load. When using data-driven methods to predict loads, only the weather, date type, and the time series periodicity of the load itself need to be considered. However, with the widespread access to distributed renewable energy, the fluctuation of plant power consumption caused by renewable energy generation has become an issue that cannot be ignored. The traditional system load forecasting method can no longer meet the system load forecasting accuracy requirements under the new situation.

[0004] Therefore, it is urgent to analyze the impact of renewable energy access on system load forecasting and propose a method to improve the load forecasting accuracy that is suitable for wide-area distributed renewable energy access. Summary of the invention

[0005] Purpose of the invention: By considering the impact of wide-area distributed new energy access on system load, the accuracy of system load forecasting can be improved. This method combines two technical routes, data-driven and physical model, and adds processing of plant power changes caused by distributed new energy access in system load forecasting, improving the accuracy of system load forecasting in wide-area distributed new energy access scenarios, and providing accurate and reliable boundary data for scheduling operation and market optimization.

[0006] Technical solution: A method for improving load forecasting accuracy considering the access of new energy sources, comprising the following steps:

[0007] Create a sample data set;

[0008] The data of the sample data set is used to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load, and the direct prediction result A of the whole network load is predicted by the best direct prediction model of the whole network load;

[0009] Combine the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load forecasting model to obtain the optimal power load forecasting model, and use the optimal power load forecasting model to predict the indirect forecast result B of the whole network load;

[0010] Based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load, the optimal prediction result is obtained.

[0011] Furthermore, the sample data set includes:

[0012] Obtain historical system load data from the regional power grid system’s collection system;

[0013] Obtain historical distributed photovoltaic output data from the collection system of the regional power grid system;

[0014] According to the geographical location of the forecast object, the historical meteorological data of the corresponding area is obtained from the meteorological service interface of the regional power grid system, and the historical meteorological data includes temperature, humidity, radiation, wind speed, and precipitation;

[0015] According to the geographical location of the forecast object, the forecast meteorological data of the corresponding area on the forecast day is obtained from the meteorological service interface of the regional power grid system, including temperature, humidity, radiation, wind speed, and precipitation;

[0016] The distributed photovoltaic output forecast value for the forecast day is obtained from the new energy forecast application of the regional power grid system.

[0017] Furthermore, the data of the sample data set is used to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load, specifically:

[0018] A set of sample data obtained from the sample data set includes historical system load data, historical distributed photovoltaic output data, and historical meteorological data. The date of any set of sample data is T, and the system load data on the Tth day is the output feature Y T , with system load data from T-7 to T-1, historical distributed photovoltaic output data on day T, historical meteorological data from T-2 to day T, and week type on day T as input feature X T , construct a sample set, and divide the sample set into a training set and a test set; build a direct network load prediction model based on a machine learning algorithm, adjust parameters, train, test and evaluate the direct network load prediction model based on the training set and the test set, and select the one with the best test effect as the best direct network load prediction model to save.

[0019] Furthermore, the direct prediction result A of the whole network load is predicted by the best direct prediction model of the whole network load, specifically: loading the best direct prediction model of the whole network load, and changing the input feature X of the prediction day F FInput the best direct forecasting model for the entire network load and obtain the output characteristics Y of the forecast day F F , that is, the system load data forecast value L on forecast day F f,A , that is, the direct prediction result A of the whole network load; the input feature X F Including system load data from F-7 to F-1, distributed photovoltaic output forecast value on F day, historical meteorological data from F-2 to F-1, forecast meteorological data on F day, and week type on F day.

[0020] Furthermore, the optimal power load forecasting model is obtained by training the power load forecasting model by combining the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant, specifically:

[0021] Obtain historical system load data, historical distributed photovoltaic output data, and historical meteorological data from the sample data set;

[0022] Retrieve the power consumption rates of thermal power plants and new energy plants in the areas under manual maintenance;

[0023] All installed capacity except for new energy plants is simplified as thermal power plants. Based on historical system load data, historical distributed photovoltaic output data, and combined with the following formula, the plant power consumption P is calculated. fcy ,

[0024] P fcy =(LP pv )×ρ a +P pv ×ρ b

[0025] The historical system load data is recorded as L, and the historical distributed photovoltaic output data is recorded as P pv The power consumption rate of thermal power plant is ρ a , the electricity consumption rate of the new energy plant is ρ b ;

[0026] The system load data L' after deducting the factory power consumption is calculated according to the following formula:

[0027] L′=LP fcy

[0028] The date of any group of samples is T, the output feature is the system load data L' after deducting the factory power consumption on the Tth day, the input features are the system load data L' after deducting the factory power consumption from T-7 to T-1, the historical meteorological data from T-2 to T, and the week type on the Tth day, and a sample set is constructed; the sample set is divided into a training set and a test set; a power load forecasting model is built based on a machine learning algorithm, and the power load forecasting model is adjusted, trained, tested and evaluated based on the training set and the test set, and the best test result is selected as the best power load forecasting model for storage.

[0029] Furthermore, the indirect prediction result B of the whole network load predicted by the optimal power load prediction model is specifically:

[0030] Based on the optimal power load forecasting model, the system load data L' after deducting the factory power consumption is predicted to obtain the prediction result L' f , the system load forecast result L is restored according to the following formula f,B , that is, the indirect prediction result B of the whole network load,

[0031]

[0032] Where P pv,f is the predicted value of distributed photovoltaic output.

[0033] Further, based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load, the optimal prediction result is obtained, including the following steps:

[0034] The average prediction errors of the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load in a set period of time are recorded as E A 、E B ;

[0035] The weighted weight α of the direct prediction result A of the entire network load is calculated according to the following formula: A , the weighted weight α of the indirect prediction result B of the whole network load B ;

[0036]

[0037]

[0038] The direct prediction result A of the whole network load and the indirect prediction result B of the whole network load are weighted to obtain the optimal prediction result.

[0039] Furthermore, the cross-validation method is used to divide the sample set into a training set and a test set.

[0040] Furthermore, the machine learning algorithm is an XGBoost algorithm.

[0041] A system for improving load prediction accuracy considering the access of new energy sources, comprising a sample data set module, a direct prediction model module for the entire network load, a power load prediction model module and a comprehensive analysis module;

[0042] The sample data set module establishes a sample data set;

[0043] The whole network load direct prediction model module uses the data of the sample data set to train the whole network load direct prediction model to obtain the best whole network load direct prediction model, and predicts the whole network load direct prediction result A through the best whole network load direct prediction model;

[0044] The power load prediction model module combines the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load prediction model to obtain the optimal power load prediction model, and predicts the indirect prediction result B of the whole network load through the optimal power load prediction model;

[0045] The comprehensive analysis module combines the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load to obtain the optimal prediction result.

[0046] A device for improving the accuracy of load forecasting considering the access of new energy sources, comprising: one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing a method for improving the accuracy of load forecasting considering the access of new energy sources.

[0047] Beneficial effect: The load forecasting accuracy improvement method proposed in the present invention takes into account the impact of changes in distributed renewable energy power generation on system load. Compared with traditional system load forecasting that only considers the load's own timing, date type, meteorological conditions, etc., the influencing factors considered in the present invention are more complete.

[0048] The load forecasting accuracy improvement method proposed in the present invention aims at the problem of plant power consumption changes caused by distributed renewable energy power generation. Compared with the method that relies on manual experience to estimate the plant power consumption change value and correct the system load forecasting result, the present invention handles the plant power consumption problem in the system load forecasting more rigorously and precisely.

[0049] The load forecasting accuracy improvement method proposed in the present invention integrates the results of two technical ideas, data-driven and physical model, avoiding the problem of unstable prediction results of a single method; and the prediction result combination mechanism adopted by the present invention assigns different weights to the prediction results of the two ideas according to the historical error statistics, avoiding the subjectivity and limitations of artificially specified weights. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 The figure is a flow chart of a method for improving load forecasting accuracy considering the access of new energy sources. DETAILED DESCRIPTION

[0051] The technical solution of the present invention is further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0052] Embodiment 1:

[0053] like Figure 1 As shown, the present invention provides a method for improving load forecasting accuracy considering the access of new energy sources, which specifically includes:

[0054] (1) Obtaining a sample data set includes the following steps:

[0055] (1.1) Obtain historical system load data for the past year from the acquisition system, with a data granularity of 15 minutes;

[0056] (1.2) Obtain the historical distributed photovoltaic output data of the past year from the collection system, with a data granularity of 15 minutes;

[0057] (1.3) Obtain the historical meteorological data of the forecast object in the past year from the meteorological service interface according to the geographical location of the forecast object, including temperature, humidity, radiation, wind speed, precipitation, etc., with a data granularity of 15 minutes;

[0058] (1.4) Obtain the forecast meteorological data of the forecast object on the forecast day from the meteorological service interface according to the geographical location of the forecast object, including temperature, humidity, radiation, wind speed, precipitation, etc., with a data granularity of 15 minutes;

[0059] (1.5) Obtain the distributed photovoltaic output forecast value for the forecast day from the new energy forecast application, with a data granularity of 15 minutes.

[0060] (2) Taking the distributed photovoltaic output as one of the input features, an artificial intelligence load forecasting model is built to obtain the load forecast result A of the entire network, including the following steps:

[0061] (2.1) Obtain the historical system load data, historical distributed photovoltaic output data, historical meteorological data of the past year from the sample data set, as well as the forecast meteorological data and distributed photovoltaic output forecast value of the forecast day;

[0062] (2.2) Let the date of any set of samples be T, and the system load data on the Tth day be the output feature Y T , with system load data from T-7 to T-1, historical distributed photovoltaic output data on day T, historical meteorological data from T-2 to day T, and week type on day T as input feature X T , construct a sample set;

[0063] (2.3) Use cross-validation method to divide the sample set into training set and test set;

[0064] (2.4) Build a prediction model based on the XGBoost algorithm, adjust the parameters of the prediction model, train, test and evaluate the model effect based on the training set and test set, and select the prediction model with the best test effect to save;

[0065] (2.5) Load the prediction model and convert the input feature X of the prediction day F into F , that is, the system load data from F-7 to F-1, the distributed photovoltaic output forecast value on F, the meteorological data from F-2 to F (F-2 to F-1 are historical meteorological data, and F is forecast meteorological data), and the week type on F are input into the prediction model to predict the output feature Y of the prediction day F F , that is, the system load data forecast value A of F, denoted as L f,A .

[0066] (3) Restore the real load demand based on the distributed photovoltaic output, photovoltaic power consumption rate, and thermal power plant power consumption rate, predict the real load demand, calculate and superimpose the power consumption to obtain the indirect prediction result B of the whole network load, which is characterized by comprising the following steps:

[0067] (3.1) Obtain historical load data, historical distributed photovoltaic output data, historical meteorological data for the past year from the sample set, as well as forecast meteorological data and distributed photovoltaic output forecast values ​​for the forecast day.

[0068] (3.2) Manually maintain the power consumption rate of regional thermal power plants and new energy power plants.

[0069] (3.3) Considering that thermal power and renewable energy account for the largest proportion in actual scenarios, in order to ensure the operability of the solution, all installed capacity except renewable energy is simplified to thermal power. Based on historical system load data, historical distributed photovoltaic output data, and combined with the following formula, the plant power P is calculated. fcy The historical system load data is recorded as L, and the historical distributed photovoltaic output data is recorded as P pv The power consumption rate of thermal power plant is ρ a , the electricity consumption rate of the new energy plant is ρ b .

[0070] P fcy =(LP pv )×ρ a +P pv ×ρ b

[0071] (3.4) The system load L' after deducting the factory power consumption is calculated according to the following formula.

[0072] L′=LP fcy

[0073] (3.5) Based on the XGBoost algorithm, the system load L' after deducting the factory power consumption is modeled and predicted to obtain the prediction result L' f The date of any group of samples is T, the output feature is the system load data after deducting the factory power consumption on the Tth day, the input feature is the system load data after deducting the factory power consumption from T-7 to T-1, the historical meteorological data from T-2 to T, and the week type on the Tth day. The training and prediction process is the same as (2.2) to (2.5).

[0074] (3.6) The system load forecast result L is restored according to the following formula f,B , which is the indirect prediction result B of the whole network load. In the formula, P pv,f is the predicted value of distributed photovoltaic output.

[0075]

[0076] (4) Combining the prediction results A and B to obtain the optimal prediction result includes the following steps:

[0077] (4.1) The average prediction errors of prediction results A and B in the last 30 days are denoted as E A 、E B ;

[0078] (4.2) Calculate the weighted weight α of the two prediction results according to the following formula: A , α B ;

[0079]

[0080]

[0081] (4.3) The optimal prediction result L is obtained by weighting according to the following formula f .

[0082] L f =α A L f,A +α B L f,B

[0083] The present invention aims to solve the problem that wide-area distributed new energy access exacerbates system load volatility and prediction difficulty, and adds processing of plant power changes caused by distributed new energy access in system load prediction. It combines the data-driven method of adding distributed new energy output data to the system load prediction model, and the physical model method of decoupling the prediction of plant power and power load and then restoring it to system load prediction, so as to improve the system load prediction accuracy in the wide-area distributed new energy access scenario, and provide accurate and reliable boundary data for scheduling operation and market optimization.

[0084] Embodiment 2:

[0085] A system for improving load prediction accuracy considering the access of new energy sources, comprising a sample data set module, a direct prediction model module for the entire network load, a power load prediction model module and a comprehensive analysis module;

[0086] The sample data set module establishes a sample data set;

[0087] The whole network load direct prediction model module uses the data of the sample data set to train the whole network load direct prediction model to obtain the best whole network load direct prediction model, and predicts the whole network load direct prediction result A through the best whole network load direct prediction model;

[0088] The power load prediction model module combines the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load prediction model to obtain the optimal power load prediction model, and predicts the indirect prediction result B of the whole network load through the optimal power load prediction model;

[0089] The comprehensive analysis module obtains the optimal prediction result based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load.

[0090] Furthermore, the sample data set includes:

[0091] Obtain historical system load data from the regional power grid system’s collection system;

[0092] Obtain historical distributed photovoltaic output data from the collection system of the regional power grid system;

[0093] According to the geographical location of the forecast object, historical meteorological data of the corresponding area is obtained from the meteorological service interface of the regional power grid system, wherein the historical meteorological data includes temperature, humidity, radiation, wind speed, and precipitation;

[0094] According to the geographical location of the forecast object, the forecast meteorological data of the corresponding area on the forecast day is obtained from the meteorological service interface of the regional power grid system, including temperature, humidity, radiation, wind speed, and precipitation;

[0095] The distributed photovoltaic output forecast value for the forecast day is obtained from the new energy forecast application of the regional power grid system.

[0096] Furthermore, the data of the sample data set is used to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load, specifically:

[0097] Obtain historical system load data, historical distributed photovoltaic output data, historical meteorological data, as well as forecast meteorological data and distributed photovoltaic output forecast values ​​on the forecast day from the sample data set; denote the date of any set of sample data as T, and take the system load data on the Tth day as the output feature Y T , with system load data from T-7 to T-1, historical distributed photovoltaic output data on day T, historical meteorological data from T-2 to day T, and week type on day T as input feature X T , construct a sample set, and divide the sample set into a training set and a test set; the week types include weeks 1-7; build a direct network load prediction model based on a machine learning algorithm, adjust parameters, train, test and evaluate the direct network load prediction model based on the training set and the test set, and select the one with the best test effect as the best direct network load prediction model for storage.

[0098] Furthermore, the method of predicting the direct prediction result A of the entire network load by using the best direct prediction model of the entire network load is specifically as follows: loading the best direct prediction model of the entire network load, and converting the input feature X of the prediction day F F Input the best direct forecasting model for the entire network load and obtain the output characteristics Y of the forecast day F F , that is, the system load data forecast value A on forecast day F, denoted as L f,A ; The input feature X F Including system load data from F-7 to F-1, distributed photovoltaic output forecast value on F day, historical meteorological data from F-2 to F-1, forecast meteorological data on F day, and week type on F day.

[0099] Furthermore, the optimal power load forecasting model is obtained by training the power load forecasting model by combining the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant, specifically:

[0100] Obtain historical system load data, historical distributed photovoltaic output data, historical meteorological data, as well as forecast meteorological data and distributed photovoltaic output forecast values ​​on the forecast day from the sample data set;

[0101] Retrieve the power consumption rates of thermal power plants and new energy plants in the areas under manual maintenance;

[0102] All installed capacity except for new energy plants is simplified as thermal power plants. Based on historical system load data, historical distributed photovoltaic output data, and combined with the following formula, the plant power consumption P is calculated. fcy ,

[0103] P fcy =(LP pv )×ρ a +P pv ×ρ b

[0104] The historical system load data is recorded as L, and the historical distributed photovoltaic output data is recorded as P pv The power consumption rate of thermal power plant is ρ a , the electricity consumption rate of the new energy plant is ρ b ;

[0105] The system load data L' after deducting the factory power consumption is calculated according to the following formula:

[0106] L′=LP fcy

[0107] The date of any group of samples is T, the output feature is the system load data after deducting the factory power consumption on the Tth day, the input features are the system load data after deducting the factory power consumption from T-7 to T-1, the historical meteorological data from T-2 to T, and the week type on the Tth day, and a sample set is constructed; the sample set is divided into a training set and a test set; a power load forecasting model is built based on a machine learning algorithm, and the power load forecasting model is adjusted, trained, tested and evaluated based on the training set and the test set, and the best test result is selected as the best power load forecasting model for storage.

[0108] Furthermore, the indirect prediction result B of the whole network load predicted by the optimal power load prediction model is specifically:

[0109] Based on the optimal power load forecasting model, the system load data L' after deducting the factory power consumption is predicted to obtain the prediction result L' f , the system load forecast result L is restored according to the following formula f,B , that is, the indirect prediction result B of the whole network load,

[0110]

[0111] Where P pv,f is the predicted value of distributed photovoltaic output.

[0112] Furthermore, the cross-validation method is used to divide the sample set into a training set and a test set.

[0113] Furthermore, the machine learning algorithm is an XGBoost algorithm.

[0114] Embodiment 3:

[0115] A device for improving the accuracy of load forecasting considering the access of new energy sources, comprising: one or more processors, one or more memories and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing a method for improving the accuracy of load forecasting considering the access of new energy sources.

[0116] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

[0117] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0118] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute the lightweight power load forecasting method as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0119] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present application (including the claims) is limited to these examples. In line with the concept of the present application, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present application as described above, which are not provided in detail for the sake of simplicity.

[0120] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present application difficult to understand, the known power supply / ground connection with the integrated circuit (IC) chip and other components may or may not be shown in the provided drawings. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present application difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform to be implemented in the embodiments of the present application (that is, these details should be fully within the scope of understanding of those skilled in the art). In the case of elaborating specific details (e.g., circuits) to describe exemplary embodiments of the present application, it is obvious to those skilled in the art that the embodiments of the present application can be implemented without these specific details or when these specific details are changed. Therefore, these descriptions should be considered to be illustrative rather than restrictive.

[0121] Although the present application has been described in conjunction with specific embodiments of the present application, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

Claims

1. A method for improving load prediction accuracy considering the access of new energy sources, characterized in that: The following steps are involved: Create a sample data set; The data of the sample data set is used to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load, and the direct prediction result A of the whole network load is predicted by the best direct prediction model of the whole network load; Combine the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load forecasting model to obtain the optimal power load forecasting model, and use the optimal power load forecasting model to predict the indirect forecast result B of the whole network load; Based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load, the optimal prediction result is obtained.

2. According to claim 1, a method for improving load prediction accuracy considering the access of new energy sources is characterized in that: The sample data set includes: Obtain historical system load data from the regional power grid system’s collection system; Obtain historical distributed photovoltaic output data from the collection system of the regional power grid system; Acquire historical meteorological data of the forecast object from the meteorological service interface of the regional power grid system according to the geographical location of the forecast object, wherein the historical meteorological data includes temperature, humidity, radiation, wind speed, and precipitation; Obtain the forecast meteorological data of the forecast object on the forecast day from the meteorological service interface of the regional power grid system according to the geographical location of the forecast object, including temperature, humidity, radiation, wind speed, and precipitation; In the application of new energy forecasting in regional power grid systems, the distributed photovoltaic installed capacity and irradiation forecast values ​​are used to calculate the distributed photovoltaic output forecast value on the forecast day.

3. According to claim 1, a method for improving load prediction accuracy considering the access of new energy sources is characterized in that: The method of using the data of the sample data set to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load is specifically as follows: A set of sample data obtained from the sample data set includes historical system load data, historical distributed photovoltaic output data, and historical meteorological data. The date of any set of sample data is T, and the system load data on the Tth day is the output feature Y T , with system load data from T-7 to T-1, historical distributed photovoltaic output data on day T, historical meteorological data from T-2 to day T, and week type on day T as input feature X T , construct a sample set, and divide the sample set into a training set and a test set; the week types include weeks 1-7; build a direct network load prediction model based on a machine learning algorithm, adjust parameters, train, test and evaluate the direct network load prediction model based on the training set and the test set, and select the one with the best test effect as the best direct network load prediction model for storage.

4. According to claim 1, a method for improving load prediction accuracy considering the access of new energy sources is characterized in that: The method of predicting the direct prediction result A of the entire network load by using the best direct prediction model of the entire network load is as follows: loading the best direct prediction model of the entire network load, and converting the input feature X of the prediction day F into F Input the best direct forecasting model for the entire network load and obtain the output characteristics Y of the forecast day F F , that is, the system load data forecast value L on forecast day F f,A , that is, the direct prediction result A of the whole network load; the input feature X F Including system load data from F-7 to F-1, distributed photovoltaic output forecast value on F day, historical meteorological data from F-2 to F-1, forecast meteorological data on F day, and week type on F day.

5. According to claim 1, a method for improving load prediction accuracy considering new energy access is characterized in that: The optimal power load forecasting model is obtained by combining the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load forecasting model. Specifically, Obtain historical system load data, historical distributed photovoltaic output data, and historical meteorological data from the sample data set; Retrieve the power consumption rates of thermal power plants and new energy plants in the areas under manual maintenance; All installed capacity except for new energy plants is simplified as thermal power plants. Based on historical system load data, historical distributed photovoltaic output data, and combined with the following formula, the plant power consumption P is calculated. fcy , P fcy =(L-P pv )×ρ a +P pv ×ρ b The historical system load data is recorded as L, and the historical distributed photovoltaic output data is recorded as P pv The power consumption rate of thermal power plant is ρ a , the electricity consumption rate of the new energy plant is ρ b ; The system load data L' after deducting the factory power consumption is calculated according to the following formula: L′=L-P fcy The date of any group of samples is T, the output feature is the system load data L' after deducting the factory power consumption on the Tth day, the input features are the system load data L' after deducting the factory power consumption from T-7 to T-1, the historical meteorological data from T-2 to T, and the week type on the Tth day, and a sample set is constructed; the sample set is divided into a training set and a test set; a power load forecasting model is built based on a machine learning algorithm, and the power load forecasting model is adjusted, trained, tested and evaluated based on the training set and the test set, and the best test result is selected as the best power load forecasting model for storage.

6. A method for improving load prediction accuracy considering new energy access according to claim 5, characterized in that: The indirect prediction result B of the whole network load predicted by the optimal power load prediction model is specifically: Based on the optimal power load forecasting model, the system load data L' after deducting the factory power consumption is predicted to obtain the prediction result L' f , the system load forecast result L is restored according to the following formula f,B , that is, the indirect prediction result B of the whole network load, Where P pv,f is the predicted value of distributed photovoltaic output.

7. A method for improving load prediction accuracy considering new energy access according to claim 1, characterized in that: Based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load, the optimal prediction result is obtained, including the following steps: The average prediction errors of the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load in a set period of time are recorded as E A 、E B ; The weighted weight α of the direct prediction result A of the entire network load is calculated according to the following formula: A , the weighted weight α of the indirect prediction result B of the whole network load B ; The direct prediction result A of the whole network load and the indirect prediction result B of the whole network load are weighted to obtain the optimal prediction result.

8. A system for improving load forecasting accuracy considering the access of new energy, characterized in that: It includes sample data set module, direct network load prediction model module, power load prediction model module and comprehensive analysis module; The sample data set module establishes a sample data set; The whole network load direct prediction model module uses the data of the sample data set to train the whole network load direct prediction model to obtain the best whole network load direct prediction model, and predicts the whole network load direct prediction result A through the best whole network load direct prediction model; The power load prediction model module combines the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load prediction model to obtain the optimal power load prediction model, and predicts the indirect prediction result B of the whole network load through the optimal power load prediction model; The comprehensive analysis module obtains the optimal prediction result based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load.

9. A system for improving load prediction accuracy considering new energy access according to claim 8, characterized in that: The method of using the data of the sample data set to train the direct prediction model of the whole network load to obtain the best direct prediction model of the whole network load is specifically as follows: A set of sample data obtained from the sample data set includes historical system load data, historical distributed photovoltaic output data, and historical meteorological data. The date of any set of sample data is T, and the system load data on the Tth day is the output feature Y T , with system load data from T-7 to T-1, historical distributed photovoltaic output data on day T, historical meteorological data from T-2 to day T, and week type on day T as input feature X T , construct a sample set, and divide the sample set into a training set and a test set; build a direct network load prediction model based on a machine learning algorithm, adjust parameters, train, test and evaluate the direct network load prediction model based on the training set and the test set, and select the one with the best test effect as the best direct network load prediction model to save.

10. A system for improving load prediction accuracy considering new energy access according to claim 8, characterized in that: The method of predicting the direct prediction result A of the entire network load by using the best direct prediction model of the entire network load is as follows: loading the best direct prediction model of the entire network load, and converting the input feature X of the prediction day F into F Input the best direct forecasting model for the entire network load and obtain the output characteristics Y of the forecast day F F , that is, the system load data forecast value L on forecast day F f,A , that is, the direct prediction result A of the whole network load; the input feature X F Including system load data from F-7 to F-1, distributed photovoltaic output forecast value on F day, historical meteorological data from F-2 to F-1, forecast meteorological data on F day, and week type on F day.

11. A system for improving load prediction accuracy considering new energy access according to claim 8, characterized in that: The optimal power load forecasting model is obtained by combining the data of the sample data set, the power consumption rate data of the thermal power plant and the power consumption rate data of the new energy plant to train the power load forecasting model. Specifically, Obtain historical system load data, historical distributed photovoltaic output data, and historical meteorological data from the sample data set; Retrieve the power consumption rates of thermal power plants and new energy plants in the areas under manual maintenance; All installed capacity except for new energy plants is simplified as thermal power plants. Based on historical system load data, historical distributed photovoltaic output data, and combined with the following formula, the plant power consumption P is calculated. fcy , P fcy =(L-P pv )×ρ a +P pv ×ρ b The historical system load data is recorded as L, and the historical distributed photovoltaic output data is recorded as P pv The power consumption rate of thermal power plant is ρ a , the electricity consumption rate of the new energy plant is ρ b ; The system load data L' after deducting the factory power consumption is calculated according to the following formula: L′=L-P fcy The date of any group of samples is T, the output feature is the system load data L' after deducting the factory power consumption on the Tth day, the input features are the system load data L' after deducting the factory power consumption from T-7 to T-1, the historical meteorological data from T-2 to T, and the week type on the Tth day, and a sample set is constructed; the sample set is divided into a training set and a test set; a power load forecasting model is built based on a machine learning algorithm, and the power load forecasting model is adjusted, trained, tested and evaluated based on the training set and the test set, and the best test result is selected as the best power load forecasting model for storage.

12. A system for improving load prediction accuracy considering new energy access according to claim 11, characterized in that: The indirect prediction result B of the whole network load predicted by the optimal power load prediction model is specifically: Based on the optimal power load forecasting model, the system load data L' after deducting the factory power consumption is predicted to obtain the prediction result L' f , the system load forecast result L is restored according to the following formula f,B , that is, the indirect prediction result B of the whole network load, Where P pv,f is the predicted value of distributed photovoltaic output.

13. A system for improving load prediction accuracy considering new energy access according to claim 8, characterized in that: Based on the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load, the optimal prediction result is obtained, including the following steps: The average prediction errors of the direct prediction result A of the whole network load and the indirect prediction result B of the whole network load in a set period of time are recorded as E A 、E B ; The weighted weight α of the direct prediction result A of the entire network load is calculated according to the following formula: A , the weighted weight α of the indirect prediction result B of the whole network load B ; The direct prediction result A of the whole network load and the indirect prediction result B of the whole network load are weighted to obtain the optimal prediction result.

14. A device for improving load prediction accuracy considering the access of new energy, characterized in that: include: One or more processors, one or more memories, and one or more programs, wherein the one or more programs are stored in the one or more memories and are configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the methods according to claims 1 to 6.

Images