Power prediction methods for photovoltaic energy storage systems and photovoltaic energy storage systems
By combining multi-dimensional predictive feature data and scenario correction coefficients, and utilizing technologies such as convolutional neural networks and long short-term memory networks, the coupling effect between energy storage charging and discharging and charging pile load and photovoltaic power generation in photovoltaic energy storage systems has been solved, achieving more accurate power prediction.
Patent Information
- Application Number
- CN202611601181.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-15
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies fail to effectively consider the coupling effects of energy storage charging and discharging and charging pile load with photovoltaic power generation in power prediction of photovoltaic energy storage systems, resulting in scenario adaptability defects and large power prediction errors.
By combining multi-dimensional predictive feature data, real-time energy storage status of photovoltaic energy storage system and energy storage output load status, multi-dimensional collaborative prediction and correction of photovoltaic power generation is carried out. Features are extracted using convolutional neural network, long short-term memory network and multi-head attention mechanism, and correction is performed in combination with scene correction coefficient.
It improves the accuracy and robustness of photovoltaic power generation forecasting, reduces power forecasting errors, and enhances the system's adaptability to various scenarios.
Smart Images

Figure CN123121531A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of energy storage technology, and in particular to a power prediction method for a photovoltaic energy storage system and a photovoltaic energy storage system. Background Technology
[0002] Photovoltaic energy storage systems are a new type of new energy infrastructure that integrates photovoltaic power generation, energy storage systems, and load charging. Specifically, they can include photovoltaic power generation equipment, energy storage equipment, and output load equipment such as charging piles, etc., without limitation. In the practical application of photovoltaic energy storage systems, accurate prediction of the overall output power of the system is a core prerequisite for achieving energy dispatch, power balance, and optimized absorption. Especially with the coordinated development of the new energy vehicle industry and the photovoltaic industry, the application scenarios of photovoltaic energy storage systems are continuously expanding, placing higher demands on the accuracy, real-time performance, and scenario adaptability of power prediction.
[0003] In existing technologies, when predicting the power of photovoltaic energy storage systems, a single model is often used to extract time features. The evaluation only considers the power output of the photovoltaic power generation equipment in the photovoltaic energy storage system, without taking into account the coupling effects of energy storage charging and discharging, charging pile load and photovoltaic power generation in the corresponding energy storage and charging scenarios. This results in significant scenario adaptability defects and large power prediction errors. Summary of the Invention
[0004] This application provides a power prediction method and a photovoltaic energy storage system. In the process of predicting the photovoltaic power generation of the photovoltaic energy storage system, it can combine multi-dimensional prediction feature data, real-time energy storage status of the photovoltaic energy storage system, and energy storage output load status to perform multi-dimensional collaborative prediction correction of the photovoltaic power generation prediction, which effectively improves the accuracy and robustness of photovoltaic power generation prediction.
[0005] The first aspect of this application provides a power prediction method for a photovoltaic energy storage system, which may include the following steps: determining the basic power prediction result of the photovoltaic energy storage system based on a prediction feature data set of the photovoltaic energy storage system, wherein the prediction feature data set includes one or any combination of meteorological environmental data, photovoltaic power generation data, energy storage status data, charging output data, and equipment status data of the photovoltaic energy storage system; determining the scenario correction coefficient of the photovoltaic energy storage system based on the real-time energy storage capacity data and / or real-time energy storage load data of the photovoltaic energy storage system; and determining the corrected power prediction result of the photovoltaic energy storage system based on the basic power prediction result and the scenario correction coefficient.
[0006] In one possible implementation of the first aspect above, determining the scenario correction coefficient of the photovoltaic energy storage system based on the real-time energy storage capacity data and / or real-time energy storage load data of the photovoltaic energy storage system includes the following steps: determining the real-time energy storage capacity data of the photovoltaic energy storage system based on the difference between the remaining energy storage capacity and the preset energy storage capacity; determining the real-time energy storage load data of the photovoltaic energy storage system based on the difference between the real-time energy storage load and the historical average energy storage load; and determining the scenario correction coefficient of the photovoltaic energy storage system based on the product of the real-time energy storage capacity data and the first preset correction coefficient, and the product of the real-time energy storage load data and the second preset correction coefficient.
[0007] In one possible implementation of the first aspect above, the real-time energy storage capacity data is determined based on the ratio of the difference between the remaining energy storage capacity of the photovoltaic energy storage system and the preset energy storage capacity; the real-time energy storage load data is determined based on the ratio of the difference between the real-time energy storage load of the photovoltaic energy storage system and the historical average energy storage load.
[0008] In one possible implementation of the first aspect above, determining the corrected power prediction result of the photovoltaic energy storage system based on the base power prediction result and the scenario correction coefficient includes the following steps: determining the base power prediction result corresponding to the preset prediction time range based on the preset prediction time range; and correcting the base power prediction result corresponding to the prediction time range based on the scenario correction coefficient to obtain the corrected power prediction result.
[0009] In one possible implementation of the first aspect above, determining the base power prediction result based on the prediction feature data set of the photovoltaic energy storage system includes the following steps: extracting spatial distribution features, time series features, and global dependency features from the prediction feature data set to obtain a fused feature vector corresponding to the prediction feature data set; determining the base power prediction result based on the fused feature vector using a pre-trained power prediction model; wherein, the spatial distribution features are extracted from meteorological environmental data using a convolutional neural network; the time series features are extracted from one or more of photovoltaic power generation data, energy storage status data, and charging output data using a long short-term memory network; and the global dependency features are extracted from the prediction feature data set using a multi-head attention mechanism.
[0010] In one possible implementation of the first aspect above, the power prediction method for such a photovoltaic energy storage system may further include the following steps: obtaining the actual power monitoring results of the photovoltaic energy storage system; and correcting the pre-trained power prediction model based on the difference between the actual power monitoring results and the corrected power prediction results.
[0011] In one possible implementation of the first aspect above, the power prediction method for such a photovoltaic energy storage system may further include the following steps: obtaining the actual power monitoring results of the photovoltaic energy storage system; and correcting the first preset correction coefficient and / or the second preset correction coefficient based on the difference between the actual power monitoring results and the corrected power prediction results.
[0012] In one possible implementation of the first aspect above, the power prediction method of such photovoltaic energy storage system may further include the following steps: when the difference between the actual power monitoring result and the corrected power prediction result is greater than a preset threshold, the pre-trained power prediction model is retrained based on the actual power monitoring result and the prediction feature data set.
[0013] In one possible implementation of the first aspect above, obtaining the predictive feature data set of the photovoltaic energy storage system includes: obtaining the corresponding raw acquisition data based on the data items contained in the predictive feature data; performing data normalization processing and outlier cleaning processing on the raw acquisition data to obtain the corresponding preprocessed acquisition data; and aligning the preprocessed acquisition data corresponding to the data items based on timestamp information to obtain the predictive feature data set.
[0014] The second aspect of this application provides a photovoltaic energy storage system that can perform power prediction control using the power prediction method of the photovoltaic energy storage system provided in the first aspect.
[0015] The technical solution provided in this application has at least the following advantages: The technical solution provided in this application can combine multi-dimensional prediction feature data, real-time energy storage status of photovoltaic energy storage system and energy storage output load status to perform multi-dimensional collaborative prediction correction of photovoltaic power generation prediction, effectively improving the accuracy and robustness of photovoltaic power generation prediction. Attached Figure Description
[0016] One or more embodiments are illustrated by way of example with corresponding pictures in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Unless otherwise stated, the pictures in the accompanying drawings do not constitute a limitation on scale. In order to more clearly illustrate the technical solutions in the embodiments of this application or in the conventional technology, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating a power prediction method for a photovoltaic energy storage system according to an embodiment of this application.
[0018] Figure 2This is a schematic diagram of a process for obtaining a predictive feature data set of a photovoltaic energy storage system according to an embodiment of this application.
[0019] Figure 3 This is a schematic diagram of a process for determining the prediction result of base power according to an embodiment of this application.
[0020] Figure 4 This is a flowchart illustrating a process for determining the scenario correction coefficient of a photovoltaic energy storage system according to an embodiment of this application.
[0021] Figure 5 This is a schematic flowchart illustrating a process for determining the corrected power prediction result of a photovoltaic energy storage system according to an embodiment of this application.
[0022] Figure 6 This is a schematic diagram of a process for performing power prediction iterative optimization based on the actual power monitoring results of a photovoltaic energy storage system, according to an embodiment of this application. Detailed Implementation
[0023] Based on the relevant descriptions in the background art, in the existing technology, the power prediction process for photovoltaic energy storage systems often uses a single model to extract time features and only evaluates the power output of the photovoltaic power generation equipment in the photovoltaic energy storage system. It does not consider the coupling effects of energy storage charging and discharging, charging pile load, and photovoltaic power generation in the corresponding energy storage and charging scenarios, resulting in significant scenario adaptability defects and large power prediction errors. To solve the above problems, some embodiments of this application provide a power prediction method and a photovoltaic energy storage system for photovoltaic energy storage systems. In the process of predicting the photovoltaic power generation of the photovoltaic energy storage system, it can combine multi-dimensional prediction feature data, the real-time energy storage status of the photovoltaic energy storage system, and the energy storage output load status to perform multi-dimensional collaborative prediction correction of the photovoltaic power generation prediction, effectively improving the accuracy and robustness of photovoltaic power generation prediction. The following will further explain and illustrate the specific implementation of the thermal management method for the energy storage system provided in this application with reference to embodiments.
[0024] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.
[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.
[0026] The terminology used in the description of the various embodiments herein is for the purpose of describing particular embodiments only and is not intended to be limiting. As used in the description of the various embodiments and the appended claims, the word "part" is also intended to include the plural form unless the context clearly indicates otherwise.
[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the various embodiments of this application to help readers better understand this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.
[0028] In some embodiments of this application, Figure 1 A flowchart illustrating a power prediction method for a photovoltaic energy storage system is shown; for example... Figure 1 As shown, process 100 may specifically include the following steps: Step 110: Based on the prediction feature data set of the photovoltaic energy storage system, determine the basic power prediction result of the photovoltaic energy storage system. In some embodiments, it can be understood that the photovoltaic energy storage system is a composite integrated storage and charging system that simultaneously includes a photovoltaic system and an energy storage system. It can uniformly store and manage the energy generated by the photovoltaic system and respond to energy use demand by providing local energy output and / or grid-connected energy output. Specifically, it can include photovoltaic power generation equipment, energy storage equipment, and output load equipment such as charging piles. For the photovoltaic power generation equipment, its power generation capacity is determined by the installation structure of the photovoltaic power generation equipment itself (e.g., the area of solar photovoltaic panels) and natural environmental factors (e.g., the light intensity, duration of sunlight, ambient temperature, etc. at the location). The energy storage equipment... The amount of energy stored and the output load of the output load equipment do not change the power output of the photovoltaic power generation equipment itself. However, in the photovoltaic energy storage system, the energy storage charging and discharging operation of the energy storage equipment itself and the load fluctuation of output load equipment such as charging piles will cause disturbances to the overall power output of the photovoltaic energy storage system on the site side, ultimately leading to a deviation between the actual output power of the photovoltaic energy storage system and the power output of the photovoltaic power generation equipment in the photovoltaic energy storage system. In the power prediction method provided in this application, the predicted power specifically refers to the power output of the photovoltaic system and the energy storage system together as the overall site in the photovoltaic energy storage system, which will not be elaborated on below.
[0029] In some embodiments, during the process of predicting the overall output power of the photovoltaic energy storage system at the power station level, the basic power prediction result of the photovoltaic energy storage system can be determined based on various sampled monitoring data of the photovoltaic energy storage system. The basic power prediction result corresponds to the basic predicted value of the power output of the photovoltaic system and the energy storage system together as the overall power station, which will not be elaborated further below. Specifically, in some embodiments, the prediction feature data set used in step 110 may include one or any combination of meteorological environmental data, photovoltaic power generation data, energy storage status data, charging output data, and equipment status data of the photovoltaic energy storage system. The meteorological environmental data may specifically include irradiance, ambient temperature, and ambient wind speed in the area where the photovoltaic energy storage system is located, which can be obtained based on environmental monitoring devices installed in the area where the photovoltaic energy storage system is located, and is not limited here. In some embodiments, photovoltaic power generation data may specifically include the real-time output power of photovoltaic power generation equipment in the photovoltaic energy storage system, the area of solar photovoltaic panels, the real-time surface temperature of solar photovoltaic panels, the inverter operating status, historical photovoltaic power generation curves, etc., which can be acquired based on the operating status monitoring device installed on the photovoltaic power generation equipment; in some embodiments, energy storage status data may specifically include the real-time charging and discharging power, state of charge (SOC), and state of health (SOH) of energy storage equipment in the photovoltaic energy storage system, which can be acquired based on the battery management system (BMS) in the photovoltaic energy storage system; in some embodiments, charging output data may specifically include the number of charging piles connected to the photovoltaic energy storage system, the real-time charging power of a single charging pile, the overall charging load of the photovoltaic energy storage system, and the predicted charging demand of the photovoltaic energy storage system in a future preset period; in some embodiments, equipment status data may specifically include the operating fault status information and operation and maintenance records of photovoltaic power generation equipment, energy storage equipment, output load equipment, etc., which are not limited here; users can select appropriate multi-dimensional collected feature data as part of the predictive feature data set according to the actual settings of the photovoltaic energy storage system, which are not limited here. It is understandable that the various data items in the predictive feature data set of a photovoltaic energy storage system are often based on different data sources. Different data sources have different dimensions and different collection times, so it is necessary to perform necessary data preprocessing. The specific data preprocessing process and the acquisition of the predictive feature data set will be further explained in the following text with reference to the embodiments, and will not be elaborated here.
[0030] Step 120: Based on the real-time energy storage capacity data and / or real-time energy storage load data of the photovoltaic energy storage system, determine the scenario correction coefficient of the photovoltaic energy storage system. In some embodiments, step 120 and step 110 can be executed synchronously or asynchronously, and the order of execution is not limited; the specific acquisition of the scenario correction coefficient will be explained in detail later in conjunction with the embodiments, and will not be repeated here.
[0031] Step 130: Based on the basic power prediction results and scenario correction coefficients, determine the corrected power prediction results for the photovoltaic energy storage system. The corrected power prediction results correspond to the corrected predicted power output of the photovoltaic system and the energy storage system together as a whole power station. This will not be elaborated further below. It is understood that, based on the above process 100, the power prediction method provided in this application can combine multi-dimensional prediction feature data, the real-time energy storage status of the photovoltaic energy storage system, and the energy storage output load status to perform multi-dimensional collaborative prediction correction of photovoltaic power generation, effectively improving the accuracy and robustness of photovoltaic power generation prediction. The specific implementation of the above process 100 will be further explained and illustrated below with specific embodiments.
[0032] In some embodiments, further, Figure 2 This diagram illustrates a process for obtaining a predictive feature dataset of a photovoltaic energy storage system; for example... Figure 2 As shown, process 200 may specifically include the following steps: Step 210: Based on the data items contained in the predicted feature data, obtain the corresponding raw acquisition data. In some embodiments, the acquisition of raw acquisition data can rely on various sensing facilities of the photovoltaic energy storage system. For details, please refer to the relevant descriptions in the foregoing embodiments, which will not be repeated here.
[0033] Step 220: Perform data normalization and outlier cleaning on the raw collected data to obtain the corresponding preprocessed collected data. In some embodiments, the data normalization process may specifically include using the minimum-max normalization method to map all data from all data sources to the interval [0,1] to eliminate the influence of dimensions. Specifically, this can be obtained based on the following mathematical expression: ; in, The original collected data corresponds to a single data source. The maximum extreme value in the original collected data corresponding to a single data source. The minimum extreme value in the original collected data corresponding to a single data source. This refers to the data normalization result corresponding to a single data source, and is not limited here.
[0034] In some embodiments, during the abnormal data cleaning process, an abnormal data detection model can be constructed based on the Isolation Forest algorithm. An abnormal scoring threshold (e.g., 0.8, etc., not limited here) can be set to accurately identify and block abnormal data whose scores exceed the threshold. In some embodiments, the abnormal data may be caused by sudden changes due to sensor malfunction or errors during transmission, etc., not limited here. In some embodiments, the data to be removed from the abnormal data can be completed using cubic spline interpolation to ensure the continuity and integrity of the preprocessed data, not limited here.
[0035] Step 230: Align the preprocessed collected data corresponding to the data items based on the timestamp information to obtain the prediction feature data set. In some embodiments, during the process of obtaining the prediction feature data set, all data that have completed the above data normalization processing and outlier cleaning processing can be aligned according to the unified timestamp information to construct a multi-source fusion dataset specific to the photovoltaic energy storage system, which can serve as the input basis for subsequent power prediction. This is not limited here.
[0036] In some embodiments of this application, further, Figure 3 A schematic diagram of a process for determining baseline power prediction results is shown; for example... Figure 3 As shown, process 300 may specifically include the following steps: Step 310: Based on the predicted feature data set, extract spatial distribution features, time series features, and global dependency features respectively to obtain the fused feature vector corresponding to the predicted feature data set. In some embodiments, the spatial distribution features can be extracted based on meteorological environmental data using a convolutional neural network (CNN) to characterize the distribution features of irradiance within a defined spatial grid area, which is not limited here. In other embodiments, the time series features can be extracted based on one or more of photovoltaic power generation data, energy storage status data, and charging output data using a long short-term memory network (LSTM) to characterize the temporal features of photovoltaic power generation output, energy storage status of charge changes, and charging load data changes. The time window length can be set to 24 hours, which is not limited here. In other embodiments, the global dependency features are extracted based on the predicted feature data set using a multi-head attention mechanism (Transformer), and different distribution features are weighted and fused to obtain the required fused feature vector, which is not limited here.
[0037] Step 320: Based on the fused feature vector, determine the basic power prediction result through the pre-trained power prediction model. It is understood that the technical solution provided in this application overcomes the limitations of existing photovoltaic energy storage system power prediction processes that only use a single model for time feature extraction. It can utilize convolutional neural networks to extract spatial distribution features, long short-term memory networks to capture time series features, and retain multi-head attention mechanisms to mine global long-distance dependency features, achieving multi-dimensional feature fusion and effectively improving the prediction accuracy and robustness of the power prediction model.
[0038] In some embodiments of this application, Figure 4 A flowchart illustrating the process of determining scenario correction coefficients for a photovoltaic energy storage system is shown; for example... Figure 4 As shown, process 400 may specifically include the following steps: Step 410: Based on the difference between the remaining energy storage capacity and the preset energy storage capacity of the photovoltaic energy storage system, determine the real-time energy storage capacity data of the photovoltaic energy storage system.
[0039] In some embodiments, the real-time energy storage capacity data of the photovoltaic energy storage system can be determined based on the ratio of the difference between the remaining energy storage capacity and the preset energy storage capacity of the photovoltaic energy storage system to the preset energy storage capacity; for example, the real-time energy storage capacity data of the photovoltaic energy storage system can be determined based on the following mathematical expression: ; in, This provides real-time energy storage capacity data for photovoltaic energy storage systems. The remaining energy storage capacity of the photovoltaic energy storage system is 0.5, where the maximum value of SOC is 1 and the minimum value is 0. In some embodiments, 0.5 is the reference operating point of the energy storage state of charge of the photovoltaic energy storage system. The real-time energy storage capacity data mentioned above can characterize the fluctuation range of the current state of charge of the photovoltaic energy storage system from the optimal operating range, and is not limited here.
[0040] Step 420: Based on the difference between the real-time energy storage load of the photovoltaic energy storage system and the historical average energy storage load, determine the real-time energy storage load data of the photovoltaic energy storage system. In some embodiments, step 420 and step 410 can be executed synchronously or asynchronously, and the order of execution of steps 420 and 410 is not limited.
[0041] In some embodiments, the real-time energy storage load data of the photovoltaic energy storage system can be determined based on the ratio of the difference between the real-time energy storage load and the historical average energy storage load of the photovoltaic energy storage system to the historical average energy storage load; for example, the real-time energy storage load data of the photovoltaic energy storage system can be determined based on the following mathematical expression: ; in, This provides real-time energy storage load data for photovoltaic energy storage systems. For the real-time energy storage load of the photovoltaic energy storage system, This refers to the historical average energy storage load of the photovoltaic energy storage system, which is not specified here.
[0042] Step 430: Determine the scenario correction coefficient for the photovoltaic energy storage system based on the product of real-time energy storage capacity data and a first preset correction coefficient, and the product of real-time energy storage load data and a second preset correction coefficient. In some embodiments, the scenario correction coefficient for the photovoltaic energy storage system can be determined based on the following mathematical expression: ; Wherein, is the scenario correction coefficient for the photovoltaic energy storage system, is the first preset correction coefficient, and is the second preset correction coefficient, which are not limited here. In some embodiments, the first and second preset correction coefficients can be preset according to the type of the site. For example, in community application scenarios, the load fluctuation of charging piles in the photovoltaic energy storage system is usually large. In this case, the first preset correction coefficient can be set to 0.3 and the second preset correction coefficient can be set to 0.7 to emphasize the correction effect of energy storage load data in the photovoltaic energy storage system. As another example, in industrial park application scenarios, the energy storage scheduling ratio of the photovoltaic energy storage system is relatively high. In this case, the first preset correction coefficient can be set to 0.6 and the second preset correction coefficient can be set to 0.4 to emphasize the correction effect of energy storage capacity fluctuation in the photovoltaic energy storage system, which are not limited here.
[0043] In some embodiments of this application, Figure 5 A flowchart illustrating the process of determining corrected power prediction results for a photovoltaic energy storage system is shown; for example... Figure 5 As shown, process 500 may specifically include the following steps: Step 510: Based on the preset prediction time range, determine the basic power prediction result corresponding to the prediction time range. In some embodiments of this application, the power prediction method for photovoltaic energy storage systems provided in this application further supports power prediction output at different time scales. Specifically, it can adaptively switch between short-term (e.g., any time scale between 1 hour and 24 hours) or ultra-short-term (e.g., any time scale between 15 minutes and 1 hour) output power predictions according to the energy dispatching needs of the photovoltaic energy storage station, and synchronize them to the energy management system (EMS) of the photovoltaic energy storage system to provide accurate output power prediction support for real-time power smoothing and next-day planned dispatching in actual application scenarios. This is not limited here.
[0044] Step 520: Based on the scene correction coefficient, correct the base power prediction result corresponding to the prediction time range to obtain the corrected power prediction result. In some embodiments, the corrected power prediction result can be obtained based on the following mathematical expression: ; in, The corrected power prediction results for the photovoltaic energy storage system. This provides the baseline power prediction results for photovoltaic energy storage systems. The scenario correction factor for the photovoltaic energy storage system is not limited here. In some embodiments, exemplarily taking a photovoltaic energy storage system in a community application scenario as an example, the first preset correction factor is set to 0.3, the second preset correction factor is set to 0.7, the real-time state of charge of the photovoltaic energy storage system is 0.7, and the real-time energy storage load of the photovoltaic energy storage system is higher than the historical average energy storage load. The calculated scenario correction factor... The value is 0.12. With the base power prediction result of 800kW for the photovoltaic energy storage system, the corrected power prediction result of the photovoltaic energy storage system is 896kW, and the actual external output power of the photovoltaic energy storage system is 882kW. Compared with the uncorrected base power prediction result, the corrected power prediction result of the photovoltaic energy storage system is closer to the actual output power of the photovoltaic energy storage system, thus improving the accuracy of power prediction.
[0045] It is understood that, based on the descriptions of the foregoing embodiments, the technical solution provided by this application can support accurate prediction of the power of a photovoltaic energy storage system. To further improve the accuracy of power prediction, the technical solution provided by this application also supports further iteration, correction, and optimization of the power prediction method provided in the foregoing embodiments based on the real-time power detection results of the photovoltaic energy storage system. In some embodiments, exemplarily, Figure 6 This diagram illustrates a process for iterative optimization of power prediction based on actual power monitoring results of a photovoltaic energy storage system; as shown below. Figure 6 As shown, process 600 may specifically include the following steps: Step 610: Obtain the actual power monitoring results of the photovoltaic energy storage system. In some embodiments, the actual output power of the photovoltaic energy storage system can be continuously monitored and obtained based on the operating status monitoring equipment of the photovoltaic energy storage system. This provides a basis for comparison data for the verification of the power prediction method of this application and for subsequent iterative optimization of the power prediction of the photovoltaic energy storage system under different operating scenarios / conditions. This is not limited here.
[0046] Step 620: Based on the difference between the actual power monitoring results and the corrected power prediction results, revise the pre-trained power prediction model. It is understood that during the actual operation of the photovoltaic energy storage system, the model weight parameters of the power prediction model can be updated offline based on the daily newly added photovoltaic energy storage system operation data to continuously improve the generalization ability of the power prediction model; this is not limited here.
[0047] In some embodiments of this application, further, such as Figure 6 The illustrated process 600 may further include step 630: adjusting the first preset correction coefficient and / or the second preset correction coefficient based on the difference between the actual power monitoring result and the corrected power prediction result. In some embodiments, step 630 may be performed after step 610 is completed, and may be performed either or simultaneously with step 620 provided in the foregoing embodiments, without limitation. In some embodiments, in addition to offline fine-tuning of the weight parameters of the power prediction model according to the prediction error, online fine-tuning of the first preset correction coefficient and / or the second preset correction coefficient required for calculating the scenario correction coefficient may also be performed according to the prediction error, which can also continuously improve the generalization ability of the power prediction model, without limitation.
[0048] In some embodiments of this application, further, such as Figure 6 The illustrated process 600 may further include step 640: when the difference between the actual power monitoring result and the corrected power prediction result is greater than a preset threshold, retraining the pre-trained power prediction model based on the actual power monitoring result and the prediction feature data set. In some embodiments, step 640 may be performed after step 610 is completed, and may be performed either or simultaneously with steps 620 and 630 provided in the foregoing embodiments, without limitation. In some embodiments, it is understood that when the difference between the actual power monitoring result and the corrected power prediction result is greater than a preset threshold, for example, when the deviation between the corrected power prediction result and the actual power monitoring result is greater than 5% or 10%, it can be determined that the corrected power prediction result still has a large deviation from the actual power monitoring result. In this case, retraining the pre-trained power prediction model can be performed based on the actual power monitoring result and the prediction feature data set to comprehensively retrain, verify, and adjust the various weight parameters in the power prediction model, so that the prediction results provided by the retrained power prediction model can better match the actual operation of the photovoltaic energy storage system, without limitation.
[0049] In some embodiments of this application, a photovoltaic energy storage system is also provided. This photovoltaic energy storage system can perform power prediction control by applying the power prediction method of the photovoltaic energy storage system provided in the foregoing embodiments, which will not be elaborated here.
[0050] Those skilled in the art will understand that the above embodiments are specific examples of implementing this application, and in practical applications, various changes in form and detail can be made without departing from the spirit and scope of this application. Any person skilled in the art can make various alterations and modifications without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.
Claims
1. A power prediction method for a photovoltaic energy storage system, characterized in that, include: Based on the predicted feature data set of the photovoltaic energy storage system, the basic power prediction result of the photovoltaic energy storage system is determined. The predicted feature data set includes one or any combination of meteorological environment data, photovoltaic power generation data, energy storage status data, charging output data and equipment status data of the photovoltaic energy storage system. Based on the real-time energy storage capacity data and / or real-time energy storage load data of the photovoltaic energy storage system, determine the scenario correction coefficient of the photovoltaic energy storage system; Based on the basic power prediction results and the scenario correction coefficients, the corrected power prediction results of the photovoltaic energy storage system are determined.
2. The power prediction method for a photovoltaic energy storage system according to claim 1, characterized in that, The determination of the scenario correction coefficient for the photovoltaic energy storage system based on the real-time energy storage capacity data and / or real-time energy storage load data includes: Based on the difference between the remaining energy storage capacity and the preset energy storage capacity of the photovoltaic energy storage system, the real-time energy storage capacity data of the photovoltaic energy storage system is determined; Based on the difference between the real-time energy storage load of the photovoltaic energy storage system and the historical average energy storage load, the real-time energy storage load data of the photovoltaic energy storage system is determined. The scenario correction coefficient of the photovoltaic energy storage system is determined based on the product of the real-time energy storage capacity data and the first preset correction coefficient, and the product of the real-time energy storage load data and the second preset correction coefficient.
3. The power prediction method for a photovoltaic energy storage system according to claim 2, characterized in that, The real-time energy storage capacity data is determined based on the ratio of the difference between the remaining energy storage capacity of the photovoltaic energy storage system and the preset energy storage capacity to the preset energy storage capacity. The real-time energy storage load data is determined based on the difference between the real-time energy storage load of the photovoltaic energy storage system and the historical average energy storage load, and the ratio of the difference to the historical average energy storage load.
4. The power prediction method for a photovoltaic energy storage system according to claim 1 or 2, characterized in that, The step of determining the corrected power prediction result of the photovoltaic energy storage system based on the base power prediction result and the scenario correction coefficient includes: Based on a preset prediction time range, determine the base power prediction result corresponding to the prediction time range; Based on the scenario correction coefficient, the base power prediction result corresponding to the prediction time range is corrected to obtain the corrected power prediction result.
5. The power prediction method for a photovoltaic energy storage system according to claim 1, characterized in that, The determination of the base power prediction result based on the prediction feature data set of the photovoltaic energy storage system includes: Based on the predicted feature data set, spatial distribution features, time series features, and global dependency features are extracted respectively to obtain the fused feature vector corresponding to the predicted feature data set. Based on the fused feature vector, the basic power prediction result is determined through a pre-trained power prediction model; The spatial distribution features are obtained by extracting them from the meteorological environment data through a convolutional neural network. The time series features are extracted from one or more of the photovoltaic power generation data, the energy storage status data, and the charging output data through a long short-term memory network. The global dependency features are extracted based on the predicted feature data set through a multi-head attention mechanism.
6. The power prediction method for a photovoltaic energy storage system according to claim 5, characterized in that, Also includes: Obtain the actual power monitoring results of the photovoltaic energy storage system; Based on the difference between the actual power monitoring results and the corrected power prediction results, the pre-trained power prediction model is corrected.
7. The power prediction method for a photovoltaic energy storage system according to claim 2, characterized in that, Also includes: Obtain the actual power monitoring results of the photovoltaic energy storage system; Based on the difference between the actual power monitoring results and the corrected power prediction results, the first preset correction coefficient and / or the second preset correction coefficient are corrected.
8. The power prediction method for a photovoltaic energy storage system according to claim 6 or 7, characterized in that, Also includes: If the difference between the actual power monitoring result and the corrected power prediction result is greater than a preset threshold, the pre-trained power prediction model is retrained based on the actual power monitoring result and the prediction feature data set.
9. The power prediction method for a photovoltaic energy storage system according to claim 1 or 5, characterized in that, The acquisition of the predictive feature data set of the photovoltaic energy storage system includes: Based on the data items contained in the predicted feature data, the corresponding raw collected data is obtained; Perform data normalization and abnormal data cleaning on the raw collected data to obtain the corresponding preprocessed collected data. The preprocessed collected data corresponding to the data items are aligned based on the timestamp information to obtain the predicted feature data set.
10. A photovoltaic energy storage system, characterized in that, Power prediction control is performed using the power prediction method for a photovoltaic energy storage system as described in any one of claims 1 to 9.
