Distributed photovoltaic-oriented intelligent power prediction method and system

Abstract

The invention provides a distributed photovoltaic-oriented intelligent power prediction method and system, and relates to the technical field of power prediction, and the method comprises the steps: carrying out the collection of historical monitoring data at a plurality of power station ends of a distributed photovoltaic power station according to a preset historical time zone, and obtaining a plurality of power station historical monitoring data sets; constructing a plurality of local power prediction models; carrying out real-time monitoring data acquisition at the plurality of power station ends to obtain a plurality of power station real-time monitoring data sets; inputting a plurality of local power prediction models, and obtaining a plurality of local power prediction results of a predetermined future time zone; and the prediction result is uploaded to a power grid dispatching center, and power grid dispatching optimization is carried out in combination with the power utilization demand information of the power utilization side. The method solves the technical problems that centralized power prediction of a traditional method is difficult to adapt to differentiated requirements of different power stations, accurate prediction cannot be carried out independently for each power station, and power prediction is poor in flexibility and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of power prediction, and in particular to an intelligent power prediction method and system for distributed photovoltaics. Background Art

[0002] Distributed photovoltaic power generation has become an important part of the modern power system due to its flexibility and low-carbon characteristics. However, photovoltaic power generation has strong uncertainty and volatility. Traditional distributed photovoltaic power generation power forecasting methods often cannot fully cope with the instability of distributed photovoltaic power generation. Specifically, multiple power stations in distributed photovoltaic power generation are usually located in different geographical locations and have very different external environmental conditions. Traditional centralized power forecasting methods are difficult to adapt to these differentiated needs and cannot perform accurate local forecasts for each power station individually. Especially in the case of large-scale grid connection, traditional power forecasting methods lack flexibility and accuracy. Summary of the invention

[0003] This application provides an intelligent power prediction method and system for distributed photovoltaics, aiming to solve the technical problem that the traditional centralized power prediction method is difficult to adapt to the differentiated needs of different power stations and cannot make accurate predictions for each power station separately, resulting in poor flexibility and accuracy of power prediction.

[0004] The first aspect disclosed in the present application provides an intelligent power prediction method for distributed photovoltaics, the method comprising: collecting historical monitoring data at multiple power station ends of a distributed photovoltaic power station according to predetermined historical time zones to obtain multiple power station historical monitoring data sets; constructing multiple local power prediction models local to the multiple power station ends based on the multiple power station historical monitoring data sets; collecting real-time monitoring data at the multiple power station ends to obtain multiple power station real-time monitoring data sets; inputting the multiple power station real-time monitoring data sets into the multiple local power prediction models to obtain multiple local power prediction results for predetermined future time zones; uploading the multiple local power prediction results to a power grid dispatching center through the multiple power station ends, and optimizing power grid dispatching in combination with the power demand information on the power consumption side.

[0005] The second aspect disclosed in the present application provides an intelligent power prediction system for distributed photovoltaics, the system is used for the above-mentioned intelligent power prediction method for distributed photovoltaics, and the system includes: a historical monitoring data acquisition module, the historical monitoring data acquisition module is used to collect historical monitoring data at multiple power station ends of the distributed photovoltaic power station according to a predetermined historical time zone, and obtain multiple power station historical monitoring data sets; a power prediction model construction module, the power prediction model construction module is used to construct multiple local power prediction models local to the multiple power station ends based on the multiple power station historical monitoring data sets; a real-time monitoring data acquisition module, the real-time monitoring data acquisition module is used to collect real-time monitoring data at the multiple power station ends, and obtain multiple power station real-time monitoring data sets; a power prediction result acquisition module, the power prediction result acquisition module is used to input the multiple power station real-time monitoring data sets into the multiple local power prediction models, and obtain multiple local power prediction results for the predetermined future time zone; a power grid dispatch optimization module, the power grid dispatch optimization module is used to upload the multiple local power prediction results to the power grid dispatch center through the multiple power station ends, and perform power grid dispatch optimization in combination with the power demand information on the power consumption side.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By collecting and analyzing the historical monitoring data of each power station, a local power prediction model specific to each power station is constructed. These models can perform adaptive learning based on the unique environmental conditions of the power station, including the external environment and internal equipment status, thereby improving the accuracy of power prediction. Real-time data is collected at each power station, and the real-time data is input into the local power prediction model of the power station for local prediction, and the respective power prediction results are output, thus realizing distributed management. This distributed processing method can be independently optimized according to the different environmental conditions of each power station, thereby improving the personalization and accuracy of the power prediction of each power station. The real-time power prediction results of multiple power stations are uploaded to the power grid dispatching center for centralized management. By collecting the future power prediction results of each power station, the dispatching center can dynamically grasp the changes in photovoltaic power generation in each future period, and perform global dispatch optimization in combination with the real-time demand information on the power consumption side. This intelligent dispatching method can balance power generation and power consumption according to actual conditions, effectively avoid the instability of the power grid, and optimize the operation efficiency of the power grid.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A schematic diagram of a flow chart of an intelligent power prediction method for distributed photovoltaics provided in an embodiment of the present application;

[0010] Figure 2 A schematic diagram of the structure of an intelligent power prediction system for distributed photovoltaics provided in an embodiment of the present application.

[0011] Explanation of the accompanying drawings: historical monitoring data collection module 10, power prediction model construction module 20, real-time monitoring data collection module 30, power prediction result acquisition module 40, power grid dispatch optimization module 50. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide an intelligent power prediction method and system for distributed photovoltaics, thereby solving the technical problem that the centralized power prediction of the traditional method is difficult to adapt to the differentiated needs of different power stations and cannot make accurate predictions for each power station separately, resulting in poor flexibility and accuracy of power prediction.

[0013] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically introduced below in conjunction with the accompanying drawings of the specification. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0014] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides an intelligent power prediction method for distributed photovoltaics, the method comprising:

[0015] According to the predetermined historical time zone, historical monitoring data is collected at multiple power station terminals of the distributed photovoltaic power station to obtain historical monitoring data sets of multiple power stations.

[0016] The predetermined historical time zone refers to the historical time range for data collection, which is determined based on the operation history of the power plant and the power forecast requirements. For example, the past year, the past few months, or a specific time period can be selected based on seasonal and meteorological conditions.

[0017] Monitoring equipment is deployed in each distributed photovoltaic power station. These devices can record various relevant parameters. The monitored data include but are not limited to the power output of the power station, external environment (such as solar irradiance, wind speed, temperature, etc.), and internal operating status (such as component temperature, inverter status, etc.). For multiple power stations, the monitoring data of each power station is stored in the local database at the power station end, forming a collection of historical monitoring data of multiple power stations.

[0018] A plurality of local power prediction models at the plurality of power station ends are constructed based on the plurality of power station historical monitoring data sets.

[0019] Extract the key factors that affect power output from historical monitoring data, that is, the input features of power prediction, including external environmental factors such as solar irradiance, cloud cover, temperature, wind speed, humidity, etc., as well as internal equipment status, such as photovoltaic module temperature, inverter status, operating efficiency, etc. Analyze the impact of external environmental factors on power output and build a power prediction channel for external factors. Similarly, analyze the impact of internal environmental factors on power output. For example, the efficiency of the inverter and the temperature of the photovoltaic module will directly affect the power output. According to the analysis results, build a power prediction channel for internal factors. Connect the power external environmental impact analysis channel and the power internal environmental impact analysis channel in parallel, that is, use their outputs as input at the same time to form a comprehensive power prediction model. Each power station end trains the corresponding power prediction model based on local data, and finally obtains multiple local power prediction models for multiple power station ends.

[0020] At the multiple power station ends, real-time monitoring data collection is performed respectively to obtain multiple power station real-time monitoring data sets.

[0021] Through the monitoring equipment deployed in each distributed photovoltaic power station, the operating status and external environment of the power station are monitored in real time. The monitoring equipment includes power monitoring equipment, environmental monitoring equipment, equipment status monitoring equipment, etc. Among them, environmental monitoring equipment such as solar irradiance sensors, thermometers, hygrometers, anemometers, etc. are used to obtain external environmental data of the power station. Equipment status monitoring equipment such as component temperature sensors, inverter efficiency monitoring equipment, etc. are used to monitor the operating status of internal equipment. According to the data collection results of the monitoring equipment, multiple power station real-time monitoring data sets are generated.

[0022] The multiple power station real-time monitoring data sets are respectively input into the multiple local power prediction models to obtain multiple local power prediction results in predetermined future time zones.

[0023] The real-time monitoring data is used as input features and input into the local power prediction model. The local power prediction model has been trained based on historical data, so the model can identify and analyze the relationship between the current real-time input features and the power output. The model predicts the power output in a specific time zone in the future by analyzing the input features. The scheduled future time zone can be a window of time in the next period of time, which is set according to the needs of power grid dispatch. For example, it can be a power output forecast for the next 5 minutes, 15 minutes, 1 hour, or even a day. The model generates the power forecast results for the corresponding time according to the scheduled future time zone, for example, the power output within the next 1 hour. Each power station has a local power prediction model and makes independent predictions based on real-time monitoring data. Finally, multiple power stations generate multiple local power prediction results.

[0024] The multiple local power prediction results are uploaded to the power grid dispatching center through the multiple power station terminals, and the power grid dispatching is optimized in combination with the power demand information on the power consumption side.

[0025] Distributed photovoltaic power stations upload power prediction results to the power grid dispatching center through the communication network. The dispatching center not only receives the power prediction results of the photovoltaic power station, but also obtains the electricity demand on the power user side. These demand information can come from different types of loads, such as residential electricity, industrial electricity, commercial electricity, etc. The power grid dispatching center compares and analyzes the power prediction results of multiple photovoltaic power stations with the demand on the power user side to ensure the balance of electricity supply and demand. For example, in a future time zone, if the predicted power output of the photovoltaic power station is not enough to meet the demand, an early warning will be issued and dispatching measures will be taken to ensure the balance of power supply and demand and stable operation of the power grid.

[0026] Furthermore, the method of constructing a plurality of local power prediction models at the plurality of power station ends based on the plurality of power station historical monitoring data sets includes:

[0027] Based on the multiple power station historical monitoring data sets, extract the first power station historical monitoring data set at the first power station end; based on the first power station historical monitoring data set, extract the first power station historical external environment monitoring data set and the first power station historical internal environment monitoring data set; based on the first power station historical external environment monitoring data set, construct a power external environment impact analysis channel; based on the first power station historical internal environment monitoring data set, construct a power internal environment impact analysis channel; connect the power external environment impact analysis channel and the power internal environment impact analysis channel in parallel to establish a first local power prediction model at the first power station end; and so on, traverse the multiple power station historical monitoring data sets to construct multiple local power prediction models at the multiple power station ends.

[0028] The first power station is a power station arbitrarily selected from the plurality of power stations as an analysis object, and a first power station historical monitoring data set of the first power station is extracted from the historical monitoring data sets of the plurality of power stations.

[0029] From the historical monitoring data set of the first power station, two types of data related to power prediction are further extracted, namely external environment monitoring data and internal environment monitoring data. These two types of data are used to analyze the external influencing factors and internal operating conditions of photovoltaic power generation respectively.

[0030] Specifically, external environment monitoring data refers to monitoring data related to the natural environment in which photovoltaic power generation is located. These data reflect the impact of external natural conditions on the power generation performance of photovoltaic power stations, such as the intensity of sunlight, which is a key factor affecting photovoltaic power generation, and ambient temperature. Temperature affects the working efficiency of photovoltaic modules, especially at high temperatures, the efficiency of modules usually decreases. Internal environment monitoring data refers to monitoring data related to the internal equipment status and operating conditions of the power station. They reflect the equipment operation of the photovoltaic power station and the impact of internal components on power output, such as photovoltaic module temperature. The operating temperature of photovoltaic modules has a significant impact on power generation efficiency, and excessive temperature will reduce power output.

[0031] Based on the historical external environment monitoring data set of the First Power Station, the impact of external environment monitoring data on photovoltaic power generation is analyzed. For example, solar irradiance is the most direct factor affecting the power output of photovoltaic power stations. As the irradiance increases, the power output usually increases. Correlation analysis methods, such as the Pearson correlation coefficient, are used to analyze the relationship between each external environment monitoring data and the power output to determine which external factors have a more significant impact on the power output. Based on the analysis results of historical external environment monitoring data, a power external environment impact analysis channel is constructed to capture the relationship between external environmental factors and power output.

[0032] Based on the historical internal environment monitoring data set of the first power station, the impact of internal environment monitoring data on photovoltaic power generation is analyzed. For example, the temperature of photovoltaic modules is an important internal factor affecting power generation efficiency. Generally, as the temperature of the modules increases, the conversion efficiency of the modules will decrease. The working state of the inverter, such as efficiency and fault conditions, directly affects the efficiency of the photovoltaic system in converting direct current to alternating current. The correlation analysis method is also used to determine the internal factors that have the most significant impact on power output. Based on the analysis results of historical internal environment monitoring data, a power internal environment impact analysis channel is constructed to capture the relationship between the operating state of internal equipment and power output.

[0033] The power external environment impact analysis channel and the power internal environment impact analysis channel are connected in parallel. Specifically, the outputs of the external environment channel and the internal environment channel are combined into the characteristic vector of the power prediction model to generate a local power prediction model at the first power station end. This is a comprehensive power prediction model built based on the external and internal environment monitoring data of the first power station, and can predict future power output based on real-time data.

[0034] By analogy, the data set of each power station is traversed, and a local power prediction model is established for each power station in the same way as the first power station. The model is deployed at the respective power station end for real-time power prediction.

[0035] Furthermore, the method of constructing a power external environment impact analysis channel based on the historical external environment monitoring data set of the first power station includes:

[0036] Based on the historical external environment monitoring data set of the first power station, a first power external environment impact factor is obtained; based on the first power external environment impact factor, a power-external environment impact correlation analysis is performed to obtain a correlation analysis result; based on the correlation analysis result, the power external environment impact analysis channel is constructed.

[0037] Extract the features that affect the power output from the historical external environment monitoring data set of the first power station. For example, through statistical analysis, select the external environment data that has a strong correlation with the power output as the external environment influencing factors. The external environment influencing factors are natural environmental factors that are directly or indirectly related to the power output of the photovoltaic power station, including solar irradiance, temperature, cloud cover, wind speed, etc.

[0038] The purpose of correlation analysis is to quantify the degree of influence of each external environmental factor on power output. Through analysis, it can be determined which external environmental factors have a significant impact on power output and which factors have a smaller impact. Specifically, the extracted external environmental influencing factors are used with the historical power output data for correlation analysis. Correlation analysis methods, such as the Pearson correlation coefficient, are used to calculate the correlation coefficient between each external environmental factor and power output. The value of the correlation coefficient ranges from -1 to 1, 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation. According to the analysis results, external environmental factors that have a strong correlation or a significant impact on power output are extracted to obtain the correlation analysis results.

[0039] According to the results of correlation analysis, select the external environmental factors that have a significant impact on power output, that is, factors with a higher correlation coefficient. For example, solar irradiance may be the factor with the greatest impact, while temperature and wind speed may be second.

[0040] According to the relationship between external environmental factors and power output, an analysis model is established to analyze the impact of the external environment on power. For example, a linear regression model is used to learn the linear relationship between external environmental factors and power output. In the analysis model, all selected external environmental influencing factors are used as parallel inputs. Each factor can independently affect the power output. The weight of each factor on the power output, that is, the degree of influence, is determined by the correlation coefficient. For example, solar irradiance accounts for the main weight and wind speed accounts for the secondary weight. The analysis model is trained using historical external environmental monitoring data and power output data. The relationship between external environmental factors and power output is learned through historical data, so that the model can accurately predict future power output and finally obtain the power external environment impact analysis channel.

[0041] Furthermore, the first external power environmental influencing factor at least includes solar irradiance, cloud cover, ambient temperature, ambient humidity, and wind speed.

[0042] The first external power environmental influencing factor includes at least solar irradiance, cloud cover, ambient temperature, ambient humidity, and wind speed. Among them, solar irradiance is the amount of solar energy received per unit area. Solar irradiance is the most critical factor affecting the power output of a photovoltaic power station. The higher the solar irradiance, the more energy the photovoltaic cells receive, and the greater the power generation is usually. Cloud cover refers to the proportion of the area covered by clouds in the sky. Cloud cover affects solar irradiance. Especially when the clouds are thick, the amount of solar energy reaching the ground will be greatly reduced, thereby reducing the power generation of the photovoltaic power station. Ambient temperature refers to the air temperature where the photovoltaic power station is located. Although the ambient temperature does not directly generate energy, it will affect the efficiency of photovoltaic modules. High temperature usually reduces the efficiency of photovoltaic modules and causes a decrease in power output. Ambient humidity refers to the water vapor content in the air. When the humidity is high, it may cause water vapor or condensation to form on the surface of the module, thereby affecting the working performance of the photovoltaic module. Wind speed refers to the speed of air flow. Wind speed can affect the working temperature of photovoltaic modules. Appropriate wind speed can help photovoltaic modules dissipate heat, especially in high temperature environments, which helps to increase power output.

[0043] Furthermore, based on the first power external environment impact factor, a power-external environment impact correlation analysis is performed to obtain a correlation analysis result, and the method includes:

[0044] The control variable method is used to perform power-external environment impact correlation analysis on the first power external environment impact factor in turn to obtain multiple correlation coefficients; the multiple correlation coefficients are sorted from high to low, and the sorting results are traversed according to a preset correlation coefficient threshold as a constraint, and the correlation coefficients are eliminated to obtain the correlation analysis result.

[0045] Using the control variable method, analyze the impact of each external environmental factor on power output one by one, including solar irradiance, cloud cover, temperature, etc. Specifically, during the analysis process, control other external environmental factors unchanged, change one of the factors one by one, so as to evaluate the correlation between the factor and power output. Use correlation analysis methods, such as the Pearson correlation coefficient, to calculate the correlation coefficient between each external environmental factor and power output. The value of the correlation coefficient ranges from -1 to 1, 1 indicates a perfect positive correlation, -1 indicates a perfect negative correlation, and 0 indicates no correlation. Each external environmental factor generates a correlation coefficient with power output, which is used to determine the influence of the factor on power output.

[0046] The external environmental factors are ranked according to their impact on power output. Generally, the higher the positive correlation coefficient, the greater the positive impact of the factor on power output; the larger the negative correlation coefficient (the larger the absolute value), the greater the negative impact of the factor on power output.

[0047] A correlation coefficient threshold is set in advance. For example, the absolute value threshold of the correlation coefficient is set to 0.3, which means that factors less than this threshold have a weak effect on power output and can be ignored. According to the set threshold, factors with a small impact on power output are eliminated. Assume that the correlation coefficient of wind speed is 0.25, which is less than the threshold of 0.3, so it can be eliminated. The sorted correlation coefficient results are traversed one by one to determine whether each external environmental factor meets the threshold requirement. Only factors that have a significant impact on power output are retained as the main input features of the subsequent power prediction model.

[0048] Furthermore, the method further comprises:

[0049] Generate a power recording node according to the predetermined future time zone; obtain a first local power prediction result of the first local power prediction model at the power recording node; when the first power station end runs to the power recording node, perform power recording to obtain a first local power operation result; compare the first local power prediction result and the first local power operation result, and perform feedback optimization of the first local power prediction model according to the comparison result.

[0050] The scheduled future time zone is a window of time in the next period of time. For example, it can be a power output forecast for the next 5 minutes, 15 minutes, 1 hour, or even a day. A power recording node is generated based on the scheduled future time zone. The power recording node refers to a specific time point in the future time zone. The power output of the power station at these time points is recorded and compared.

[0051] On the power recording node, the model is run to predict the power output at the corresponding time point and generate the first local power prediction result. For example, at the 10:05 node, the model outputs the power prediction value at 10:05 for subsequent comparison with the actual power.

[0052] The real-time monitoring system of the power station continuously monitors the power output of the power station. When the power station runs to the set power recording node, such as 10:05, the monitoring system records the actual power output of the node as the first local power operation result.

[0053] The prediction results obtained on the power recording node are compared with the actual power operation results, and the deviation between the two is calculated. If the deviation exceeds the preset threshold, the feedback optimization mechanism of the model is triggered, such as by introducing new historical data or real-time data, retraining the model and improving the prediction accuracy.

[0054] Furthermore, the method of performing feedback optimization of the first local power prediction model according to the comparison result includes:

[0055] Compare the first local power prediction result and the first local power operation result, and calculate the prediction deviation coefficient; obtain the prediction deviation coefficient threshold, and if the prediction deviation coefficient does not meet the prediction deviation coefficient threshold, generate a model optimization instruction; according to the model optimization instruction, take the prediction deviation coefficient threshold as the optimization target, activate the feedback adjustment layer embedded in the first local power prediction model to perform feedback optimization of the model.

[0056] The first local power prediction result and the first local power operation result are compared to calculate the prediction deviation coefficient, which is used to measure the error between the predicted value and the actual value. For example, the predicted power is subtracted from the actual power, and the difference is divided by the actual power to calculate the prediction deviation coefficient. This deviation coefficient reflects the difference between the model predicted power and the actual power.

[0057] The prediction deviation coefficient threshold is a standard interval to measure the prediction accuracy of the model. It can be set according to the specific requirements of the power station or the grid dispatching requirements. For example, setting the threshold to [-5%, 5%] means that the prediction results within this range of deviation coefficients are acceptable, while those exceeding this range require optimization. The deviation coefficient is judged to determine whether it is within the preset prediction deviation coefficient threshold. If it exceeds this range, a model optimization instruction is automatically generated, which is used to notify the system to adjust and optimize the model to improve the prediction accuracy.

[0058] The feedback adjustment layer is a self-optimization mechanism embedded in the power prediction model. When the model prediction accuracy does not meet the standard, the feedback adjustment layer is activated to adjust various parts of the model. The optimization goal is to adjust the prediction deviation coefficient to the preset prediction deviation coefficient threshold range, and perform feedback optimization of the model. For example, the latest actual power operation results are used as new training data to retrain the model to enhance the model's prediction ability. This feedback mechanism ensures that the model can be continuously adjusted according to the actual operation data, thereby improving the accuracy of the prediction and making the power prediction results closer to the actual operation results.

[0059] In summary, the intelligent power prediction method for distributed photovoltaics provided by the embodiment of the present application has the following technical effects:

[0060] By collecting and analyzing the historical monitoring data of each power station, a local power prediction model specific to each power station is constructed. These models can perform adaptive learning based on the unique environmental conditions of the power station, including the external environment and internal equipment status, thereby improving the accuracy of power prediction. Real-time data is collected at each power station, and the real-time data is input into the local power prediction model of the power station for local prediction, and the respective power prediction results are output, thus realizing distributed management. This distributed processing method can be independently optimized according to the different environmental conditions of each power station, thereby improving the personalization and accuracy of the power prediction of each power station. The real-time power prediction results of multiple power stations are uploaded to the power grid dispatching center for centralized management. By collecting the future power prediction results of each power station, the dispatching center can dynamically grasp the changes in photovoltaic power generation in each future period, and perform global dispatch optimization in combination with the real-time demand information on the power consumption side. This intelligent dispatching method can balance power generation and power consumption according to actual conditions, effectively avoid the instability of the power grid, and optimize the operation efficiency of the power grid.

[0061] Embodiment 2 is based on the same inventive concept as the intelligent power prediction method for distributed photovoltaics in the above embodiment. Figure 2 As shown, the embodiment of the present application provides an intelligent power prediction system for distributed photovoltaics, the system comprising:

[0062] A historical monitoring data acquisition module 10, the historical monitoring data acquisition module 10 is used to collect historical monitoring data at multiple power station ends of a distributed photovoltaic power station according to a predetermined historical time zone, and obtain multiple power station historical monitoring data sets; a power prediction model construction module 20, the power prediction model construction module 20 is used to construct multiple local power prediction models local to the multiple power station ends based on the multiple power station historical monitoring data sets; a real-time monitoring data acquisition module 30, the real-time monitoring data acquisition module 30 is used to collect real-time monitoring data at the multiple power station ends, and obtain multiple power station real-time monitoring data sets; a power prediction result acquisition module 40, the power prediction result acquisition module 40 is used to input the multiple power station real-time monitoring data sets into the multiple local power prediction models, and obtain multiple local power prediction results in a predetermined future time zone; a power grid dispatch optimization module 50, the power grid dispatch optimization module 50 is used to upload the multiple local power prediction results to the power grid dispatch center through the multiple power station ends, and perform power grid dispatch optimization in combination with the power demand information on the power consumption side.

[0063] Furthermore, the system further includes a plurality of local power prediction model building modules to perform the following operation steps:

[0064] Based on the multiple power station historical monitoring data sets, extract the first power station historical monitoring data set at the first power station end; based on the first power station historical monitoring data set, extract the first power station historical external environment monitoring data set and the first power station historical internal environment monitoring data set; based on the first power station historical external environment monitoring data set, construct a power external environment impact analysis channel; based on the first power station historical internal environment monitoring data set, construct a power internal environment impact analysis channel; connect the power external environment impact analysis channel and the power internal environment impact analysis channel in parallel to establish a first local power prediction model at the first power station end; and so on, traverse the multiple power station historical monitoring data sets to construct multiple local power prediction models at the multiple power station ends.

[0065] Furthermore, the system further includes a power external environment impact analysis channel construction module to perform the following operation steps:

[0066] Based on the historical external environment monitoring data set of the first power station, a first power external environment impact factor is obtained; based on the first power external environment impact factor, a power-external environment impact correlation analysis is performed to obtain a correlation analysis result; based on the correlation analysis result, the power external environment impact analysis channel is constructed.

[0067] Furthermore, the first external power environmental influencing factor at least includes solar irradiance, cloud cover, ambient temperature, ambient humidity, and wind speed.

[0068] Furthermore, the system further includes a correlation analysis result acquisition module to perform the following operation steps:

[0069] The control variable method is used to perform power-external environment impact correlation analysis on the first power external environment impact factor in turn to obtain multiple correlation coefficients; the multiple correlation coefficients are sorted from high to low, and the sorting results are traversed according to a preset correlation coefficient threshold as a constraint, and the correlation coefficients are eliminated to obtain the correlation analysis result.

[0070] Furthermore, the system also includes a model feedback optimization module to perform the following operation steps:

[0071] Generate a power recording node according to the predetermined future time zone; obtain a first local power prediction result of the first local power prediction model at the power recording node; when the first power station end runs to the power recording node, perform power recording to obtain a first local power operation result; compare the first local power prediction result and the first local power operation result, and perform feedback optimization of the first local power prediction model according to the comparison result.

[0072] Furthermore, the system further includes a feedback optimization module to perform the following operation steps:

[0073] Compare the first local power prediction result and the first local power operation result, and calculate the prediction deviation coefficient; obtain the prediction deviation coefficient threshold, and if the prediction deviation coefficient does not meet the prediction deviation coefficient threshold, generate a model optimization instruction; according to the model optimization instruction, take the prediction deviation coefficient threshold as the optimization target, activate the feedback adjustment layer embedded in the first local power prediction model to perform feedback optimization of the model.

[0074] Through the above-mentioned detailed description of an intelligent power prediction method for distributed photovoltaics in this specification, those skilled in the art can clearly understand an intelligent power prediction system for distributed photovoltaics in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0075] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent power prediction method for distributed photovoltaics, characterized in that: The method comprises: According to the predetermined historical time zone, historical monitoring data are collected at multiple power station terminals of the distributed photovoltaic power station to obtain a set of historical monitoring data of multiple power stations; Constructing a plurality of local power prediction models at the plurality of power station ends based on the plurality of power station historical monitoring data sets; At the multiple power station terminals, real-time monitoring data collection is performed respectively to obtain multiple power station real-time monitoring data sets; Inputting the plurality of power station real-time monitoring data sets into the plurality of local power prediction models respectively to obtain a plurality of local power prediction results for predetermined future time zones; The multiple local power prediction results are uploaded to the power grid dispatching center through the multiple power station terminals, and the power grid dispatching is optimized in combination with the power demand information on the power consumption side.

2. The intelligent power prediction method for distributed photovoltaics according to claim 1, characterized in that: The method of constructing a plurality of local power prediction models at the plurality of power station ends based on the plurality of power station historical monitoring data sets comprises: Extracting a first power station historical monitoring data set at a first power station end based on the plurality of power station historical monitoring data sets; Based on the first power station historical monitoring data set, extracting the first power station historical external environment monitoring data set and the first power station historical internal environment monitoring data set; Based on the historical external environment monitoring data set of the first power station, construct a power external environment impact analysis channel; Building a power internal environment impact analysis channel based on the historical internal environment monitoring data set of the first power station; The power external environment impact analysis channel and the power internal environment impact analysis channel are connected in parallel to establish a first local power prediction model at the first power station end; By analogy, the plurality of power station historical monitoring data sets are traversed to construct a plurality of local power prediction models at the plurality of power station ends.

3. The intelligent power prediction method for distributed photovoltaics according to claim 2, characterized in that: The method of constructing a power external environment impact analysis channel based on the historical external environment monitoring data set of the first power station includes: Based on the historical external environment monitoring data set of the first power station, obtaining a first power external environment impact factor; Based on the first power external environment impact factor, performing power-external environment impact correlation analysis to obtain a correlation analysis result; According to the correlation analysis result, the power external environment impact analysis channel is constructed.

4. The intelligent power prediction method for distributed photovoltaics according to claim 3, characterized in that: The first external power environmental influencing factor at least includes solar irradiance, cloud cover, ambient temperature, ambient humidity, and wind speed.

5. The intelligent power prediction method for distributed photovoltaics according to claim 4, characterized in that: Based on the first power external environment impact factor, a power-external environment impact correlation analysis is performed to obtain a correlation analysis result, the method comprising: Using the control variable method to sequentially perform power-external environment impact correlation analysis on the first power external environment impact factor to obtain multiple correlation coefficients; The multiple correlation coefficients are sorted from high to low, and the sorting results are traversed according to a preset correlation coefficient threshold as a constraint, and the correlation coefficients are eliminated to obtain a correlation analysis result.

6. The intelligent power prediction method for distributed photovoltaics according to claim 2, characterized in that: The method further comprises: generating a power recording node according to the predetermined future time zone; Obtain a first local power prediction result of the first local power prediction model at the power recording node; When the first power station terminal runs to the power recording node, power recording is performed to obtain a first local power operation result; The first local power prediction result and the first local power operation result are compared, and feedback optimization of the first local power prediction model is performed according to the comparison result.

7. The intelligent power prediction method for distributed photovoltaics according to claim 6, characterized in that: The method of performing feedback optimization of the first local power prediction model according to the comparison result includes: Comparing the first local power prediction result and the first local power operation result, and calculating and obtaining a prediction deviation coefficient; Obtaining a prediction deviation coefficient threshold, and generating a model optimization instruction if the prediction deviation coefficient does not meet the prediction deviation coefficient threshold; According to the model optimization instruction, taking the prediction deviation coefficient threshold as the optimization target, the feedback adjustment layer embedded in the first local power prediction model is activated to perform feedback optimization of the model.

8. An intelligent power prediction system for distributed photovoltaics, characterized in that: A system for implementing an intelligent power prediction method for distributed photovoltaics as described in any one of claims 1 to 7, the system comprising: A historical monitoring data collection module, wherein the historical monitoring data collection module is used to collect historical monitoring data at multiple power station terminals of a distributed photovoltaic power station according to a predetermined historical time zone, and obtain a set of historical monitoring data of multiple power stations; A power prediction model building module, the power prediction model building module is used to build a plurality of local power prediction models at the plurality of power station ends based on the plurality of power station historical monitoring data sets; A real-time monitoring data acquisition module, wherein the real-time monitoring data acquisition module is used to respectively perform real-time monitoring data acquisition at the multiple power station ends to obtain a set of real-time monitoring data of multiple power stations; A power prediction result acquisition module, the power prediction result acquisition module is used to input the multiple power station real-time monitoring data sets into the multiple local power prediction models respectively, and obtain multiple local power prediction results for predetermined future time zones; A power grid dispatching optimization module is used to upload the multiple local power prediction results to the power grid dispatching center through the multiple power station terminals, and perform power grid dispatching optimization in combination with the power demand information on the power consumption side.

Images