Photovoltaic power station active power optimal distribution method for short-term photovoltaic power prediction

Abstract

The invention discloses a photovoltaic power station active power optimal distribution method for short-term photovoltaic power prediction, and relates to the technical field of photovoltaic power prediction, and the method comprises the following steps: obtaining historical power generation data and real-time weather information of a photovoltaic power station, removing noise data through data preprocessing, and generating an input data set for prediction; and establishing a photovoltaic power prediction model by using a machine learning algorithm, and predicting the photovoltaic power in a future short period by using the input data set. According to the method, photovoltaic power prediction is optimized through machine learning, the prediction accuracy is improved, errors are reduced, and the accuracy of power scheduling is ensured. By means of real-time monitoring and dynamic scheduling, the risk of overload or insufficient load is avoided, and the safety and stability of power station equipment are improved. Meanwhile, through the optimized power scheduling scheme, the power generation benefit of the photovoltaic power station is maximized, efficient power generation is kept in unstable weather, the economic return of the power station is finally improved, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power prediction, and in particular to a method for optimizing active power distribution of a photovoltaic power station for short-term photovoltaic power prediction. Background Art

[0002] Optimal active power allocation in PV power plants based on short-term PV power forecasting involves rationally allocating the output power of each inverter or power generation unit within a PV plant based on the predicted PV power (i.e., the power the PV system will generate in the short term), thereby achieving optimal power scheduling and improving resource utilization efficiency. This optimization process is typically based on meteorological data, the operating status of PV panels, and historical power generation data, using an algorithmic model to predict the power generation capacity of each plant component over a period of time. This approach not only prevents overload in certain areas and equipment failures, but also maximizes the overall power generation efficiency of the plant, thereby optimizing the overall economic benefits of the plant.

[0003] Existing technologies have the following shortcomings: During the process of optimizing active power allocation in photovoltaic power plants based on short-term photovoltaic power forecasts, if the photovoltaic power forecast does not match the actual power generation, a "dynamic scheduling error" problem may occur. Because photovoltaic power generation is affected by weather changes, environmental factors, and equipment status, the predicted power often contains certain errors. If the predicted power is too high and the actual power generation is lower than expected, the power plant may over-dispatch power generation units in certain areas when optimizing power allocation, causing these units to overload and even damage inverters or other electrical equipment. Conversely, if the predicted power is too low, the power plant may miss out on some power generation potential and fail to achieve maximum power generation benefits. This problem is particularly prominent under complex meteorological conditions, such as changes in cloud cover and sudden weather events. If this occurs, it may lead to equipment failure, reduced power generation efficiency, or even system shutdown, seriously affecting the economic benefits and safe operation of the power plant.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method for optimizing active power allocation in a photovoltaic power plant based on short-term photovoltaic power forecasting. This method optimizes photovoltaic power forecasting through machine learning, improving forecast accuracy, reducing errors, and ensuring precise power scheduling. Real-time monitoring and dynamic scheduling avoid the risks of overload or underload, improving the safety and stability of power plant equipment. Furthermore, through an optimized power scheduling scheme, the power generation efficiency of the photovoltaic power plant is maximized, maintaining efficient power generation even in unstable weather conditions. Ultimately, this method improves the economic returns of the power plant and reduces operating costs, thereby addressing the aforementioned issues in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction, comprising the following steps:

[0007] Obtain historical power generation data and real-time weather information from photovoltaic power plants, remove noise data through data preprocessing, and generate input data sets for prediction;

[0008] A photovoltaic power prediction model is established using a machine learning algorithm to predict photovoltaic power in the short term using the input data set;

[0009] Analyze the deviation between the predicted power and the actual generated power, and adjust the parameters of the photovoltaic power prediction model according to the deviation to improve the prediction accuracy;

[0010] Based on the adjusted power forecast value and the operating status of each area in the power station, the power of the photovoltaic power generation unit is dispatched to ensure that the load of each power generation unit is within a safe range;

[0011] Monitor real-time power generation data and dynamically adjust the dispatch plan if power generation deviates significantly from the forecast to prevent equipment overload or underutilization of power generation potential;

[0012] By optimizing scheduling and real-time adjustments, the overall power generation efficiency of the power station can be maximized, while avoiding the risk of equipment damage and downtime caused by power scheduling errors.

[0013] Preferably, the machine learning algorithm further includes a feature engineering step, which extracts key features that affect photovoltaic power by analyzing historical power generation data and weather information of the photovoltaic power station, and through data standardization and normalization processing, makes the input data set have precise prediction accuracy. The key features include but are not limited to sunshine intensity, temperature, humidity, wind speed, seasonal changes and equipment operating status.

[0014] Preferably, the deviation analysis step further includes performing statistical analysis on the deviation of each power prediction result, and adjusting the parameters of the photovoltaic power prediction model by using a dynamic weighted average method. The weighted average method weights each deviation according to historical deviation values ​​and real-time weather changes, thereby introducing higher adaptability in the prediction process.

[0015] Preferably, the specific steps of deviation analysis are as follows:

[0016] First, the actual power generation value P at each prediction moment is obtained from the real-time monitoring system of the photovoltaic power station. actual And the predicted power value P forecast , where P actual is the actual power value obtained from the real-time measurement equipment, P forecast It is the predicted power value obtained by the photovoltaic power prediction model based on historical data and meteorological conditions;

[0017] The prediction error at each moment is calculated based on the acquired data. The calculation expression is as follows:

[0018] ε t =P forecast，t -P actual，t

[0019] , where P actual，t is the actual power generation value at time t, P forecast，t is the predicted power value at time t, ε t is the forecast error at time t;

[0020] By accumulating the prediction errors at all times, the mean absolute error is calculated as the overall performance indicator of the prediction model. The calculation expression is as follows:

[0021]

[0022] Where MAE is the mean absolute error, N is the total number of time points in the prediction period;

[0023] In addition to MAE, the root mean square error is also an important indicator to measure the accuracy of the prediction model. The root mean square error calculation expression is as follows:

[0024]

[0025] , where RMSE is the root mean square error;

[0026] The overall prediction accuracy is calculated based on MAE and RMSE. The calculation expression is as follows:

[0027]

[0028] Where ρ is the overall prediction accuracy, and β is the adjustment factor used to adjust the impact of MAE and RMSE on the overall prediction accuracy;

[0029] The prediction model is adjusted based on the overall prediction accuracy ρ. The adjustment logic is as follows:

[0030] If the overall accuracy ρ reaches the preset standard, it indicates that the prediction model performs well overall, and power scheduling can be directly performed and the optimization allocation process can be started, avoiding model adjustment and computational overhead;

[0031] If the overall pre-accuracy ρ is lower than the preset standard, further local adjustments are performed.

[0032] Preferably, power scheduling further includes real-time monitoring of the equipment health status of each power generation unit of the photovoltaic power station, and dynamically adjusting the load of the power generation unit according to the remaining life and operating efficiency of the equipment, so that the power station can maximize power output while avoiding overload operation.

[0033] Preferably, the power scheduling steps are as follows:

[0034] Define the load constraints for each photovoltaic power generation unit as follows:

[0035] L min ≤P unit，j ≤L max

[0036] , where L min is the minimum load limit of the j-th generating unit, L max is the maximum load limit of the jth generating unit, P unit，j is the actual output power of the jth power generation unit;

[0037] Set the objective function F(P unit ) is the power generation benefit function of the entire power station. The goal is to maximize the output power of the power station, and considering the efficiency of each power generation unit, the objective function formula is as follows:

[0038]

[0039] Where W j is the efficiency weight of the jth power generation unit, M is the total number of power generation units in the photovoltaic power station, F(P unit ) is the overall power generation efficiency objective function of the PV power station, which represents the sum of the products of the output power of all power generation units and their efficiency weights;

[0040] A constrained optimization algorithm is used to optimize the objective function to ensure that each power generation unit achieves maximum output power within its load constraint range while avoiding excessive load and power waste between units. The goal of the optimization algorithm is to minimize the following loss function:

[0041]

[0042] Where, P target，j is the target power of the jth power generation unit, based on the output value of the photovoltaic power prediction model, λ is the weight coefficient for load regulation and safe operation, P max，j is the maximum load limit of the j-th generating unit, It means finding P in the optimization process unit The minimum value of .

[0043] Preferably, real-time monitoring is achieved through a variety of sensors configured in the power station, including temperature sensors, light sensors, humidity sensors, and current and voltage sensors. The data collected by the sensors are transmitted wirelessly to the central control system. The central control system analyzes the operating status and environmental changes of the power station in real time based on the sensor data, and can realize error detection and dynamic adjustment between power prediction results and actual power generation, ensuring the accuracy and real-time performance of power optimization distribution.

[0044] Preferably, the optimization algorithm for maximizing power generation efficiency further includes dynamically adjusting the scheduling strategy based on the real-time weather changes of the power station and the load conditions of the power generation units to cope with sudden weather changes. The optimization algorithm takes into account the impact of weather changes on photovoltaic power generation efficiency, and through a multi-objective optimization model, while ensuring equipment safety, improves power generation efficiency, thereby improving the comprehensive benefits of the photovoltaic power station.

[0045] Preferably, real-time adjustment is performed by introducing edge computing devices, which can perform data processing and optimization calculations at the edge layer of the photovoltaic power station, reduce data transmission delays, and improve the response speed of the system; through the edge computing architecture, power scheduling can be made more flexible and responsive.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0047] This invention effectively improves the accuracy of short-term photovoltaic power forecasts. By introducing a machine learning-based photovoltaic power forecasting model and integrating it with historical power generation data and real-time weather information for refined processing, it significantly enhances the accuracy of power forecasts. Especially under complex meteorological conditions, dynamic adjustment of the model parameters reduces forecast errors, minimizing the deviation between forecast results and actual power generation capacity, thereby providing a more accurate basis for subsequent power optimization and allocation. This approach enables power plants to maintain efficient and stable power output under rapidly changing environmental conditions.

[0048] This invention effectively avoids power scheduling errors and improves power plant safety. In traditional photovoltaic power optimization and allocation, overscheduling or underloading due to prediction errors often leads to equipment damage or system downtime. Dynamic scheduling and real-time monitoring ensure that the load of each power generation unit is kept within a safe range, preventing equipment overload. Through real-time data feedback and dynamic adjustments from intelligent sensors, power plants can quickly identify anomalies and take timely action, effectively reducing the risk of equipment failure and enhancing power plant safety and reliability.

[0049] This invention significantly improves the overall power generation efficiency of photovoltaic power plants. Through an optimized power scheduling strategy, it maximizes the potential of photovoltaic power generation while avoiding equipment damage or insufficient power generation due to overscheduling. Especially in unstable weather conditions, the power plant can adjust the scheduling plan in real time to optimize load distribution based on actual power generation conditions, thereby ensuring efficient operation of the power generation units. Ultimately, the overall power generation efficiency of the power plant is significantly improved, increasing economic returns and reducing operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0051] Figure 1 The present invention provides a flow chart of a method for optimizing active power allocation in a photovoltaic power station based on short-term photovoltaic power prediction. DETAILED DESCRIPTION

[0052] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0053] The present invention provides Figure 1 The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power forecasting is shown, comprising the following steps:

[0054] Obtain historical power generation data and real-time weather information from photovoltaic power plants, remove noise data through data preprocessing, and generate input data sets for prediction;

[0055] A photovoltaic power prediction model is established using a machine learning algorithm to predict photovoltaic power in the short term using the input data set;

[0056] The machine learning algorithm further includes a feature engineering step, which extracts key features that affect photovoltaic power by analyzing historical power generation data and weather information of photovoltaic power stations, and through data standardization and normalization, the input data set has precise prediction accuracy. Key features include but are not limited to sunshine intensity, temperature, humidity, wind speed, seasonal changes and equipment operating status.

[0057] Through this feature engineering step, the adaptability of the prediction model to complex environmental factors is improved, thereby improving the accuracy of photovoltaic power prediction.

[0058] Analyze the deviation between the predicted power and the actual generated power, and adjust the parameters of the photovoltaic power prediction model according to the deviation to improve the prediction accuracy;

[0059] The deviation analysis step further includes statistical analysis of the deviations of each power forecast result, and adjusting the parameters of the photovoltaic power prediction model by using a dynamic weighted average method. The weighted average method weights each deviation based on historical deviation values ​​and real-time weather changes, thereby introducing higher adaptability in the prediction process.

[0060] It solves the problem of prediction error accumulation caused by traditional methods due to over-reliance on historical data or a single algorithm, thereby maintaining high-precision predictions under complex weather conditions.

[0061] The specific steps of deviation analysis are as follows:

[0062] First, the actual power generation value P at each prediction moment is obtained from the real-time monitoring system of the photovoltaic power station. actual And the predicted power value P forecast , where P actual is the actual power value obtained from the real-time measurement equipment, P forecast It is the predicted power value obtained by the photovoltaic power prediction model based on historical data and meteorological conditions;

[0063] The prediction error at each moment is calculated based on the acquired data. The calculation expression is as follows:

[0064] ε t =P forecast，t -P actual，t

[0065] , where P actual，t is the actual power generation value at time t, P forecast，t is the predicted power value at time t, ε t is the forecast error at time t;

[0066] By accumulating the prediction errors at all times, the mean absolute error is calculated as the overall performance indicator of the prediction model. The calculation expression is as follows:

[0067]

[0068] Where MAE is the mean absolute error, which represents the average of all forecast errors. It measures the average difference between the predicted value and the actual value. The smaller the MAE value, the more accurate the forecast. N is the total number of time points in the forecast period.

[0069] In addition to MAE, the root mean square error is also an important indicator for measuring the accuracy of the prediction model. By averaging the squares of the errors and then taking the square root, the impact of larger errors can be highlighted. The root mean square error calculation expression is as follows:

[0070]

[0071] Where RMSE is the root mean square error, which is the square root of the average of the squares of the errors. It can reflect the overall size of the error. Compared with MAE, RMSE is more sensitive to larger errors, so it is usually used to emphasize larger deviation problems.

[0072] The overall prediction accuracy is calculated based on MAE and RMSE. The calculation expression is as follows:

[0073]

[0074] Where ρ is the overall prediction accuracy, which indicates the overall accuracy of the model. A higher value indicates better prediction accuracy. α is an adjustment factor used to adjust the impact of MAE and RMSE on overall prediction accuracy. By adjusting α, we can control the weight of error on prediction accuracy. A larger α means that we place more emphasis on the impact of MAE and RMSE.

[0075] The prediction model is adjusted based on the overall prediction accuracy ρ. The adjustment logic is as follows:

[0076] If the overall accuracy ρ reaches the preset standard (e.g., ρ>0.9), it indicates that the prediction model performs well overall, and power scheduling can be directly performed and the optimization allocation process can be started, avoiding unnecessary model adjustments and computational overhead;

[0077] If the overall prediction accuracy ρ is lower than the preset standard (for example, ρ < 0.9), further local adjustments are performed. The specific steps are as follows:

[0078] For moments with large errors, local weighted adjustments are performed to calculate the prediction error at each moment and make appropriate corrections based on the historical trend of the error. For example, for moments with abnormalities, the predicted power can be fine-tuned to minimize the error.

[0079] Based on real-time meteorological data (such as cloud changes, temperature fluctuations, etc.), the model weights are redistributed, and the prediction model is adjusted by weighting historical errors and meteorological factors to make it better adapt to changes in specific environmental conditions;

[0080] During fine-tuning, the power distribution of each power generation unit is precisely adjusted according to its actual power generation situation. In this way, the load of the power generation unit can be quickly adjusted when local environmental factors fluctuate greatly, ensuring that the equipment operates within a safe range while improving power generation efficiency.

[0081] Through these steps, the prediction model can dynamically adjust and optimize power distribution under different meteorological and load conditions, ensuring that the photovoltaic power station can maintain high power generation efficiency under various environmental changes and effectively avoid the risk of equipment failure.

[0082] Based on the adjusted power forecast value and the operating status of each area in the power station, the power of the photovoltaic power generation unit is dispatched to ensure that the load of each power generation unit is within a safe range;

[0083] Power scheduling further includes real-time monitoring of the health status of each power generation unit in the photovoltaic power station, and dynamically adjusting the load of the power generation unit according to the remaining life and operating efficiency of the equipment, so that the power station can maximize power output while avoiding overload operation.

[0084] By introducing an intelligent equipment monitoring system, this scheduling method can take protective measures in advance when the equipment is about to enter the fault zone, thereby reducing the risk of equipment failure and extending the service life of the equipment.

[0085] The power scheduling steps are as follows:

[0086] Define the load constraints for each photovoltaic power generation unit as follows:

[0087] L min ≤P unit，j ≤L max

[0088] , where L min is the minimum load limit of the j-th generating unit, L max is the maximum load limit of the jth generating unit, P unit，j is the actual output power of the jth power generation unit;

[0089] Set the objective function F(P unit ) is the power generation benefit function of the entire power station. The goal is to maximize the output power of the power station, and considering the efficiency of each power generation unit, the objective function formula is as follows:

[0090]

[0091] , where W j is the efficiency weight of the jth power generation unit, M is the total number of power generation units in the photovoltaic power station, F(P unit ) is the overall power generation efficiency objective function of the PV power plant, which represents the sum of the products of the output power of all power generation units and their efficiency weights. The goal is to maximize the overall power generation efficiency while taking into account the health and efficiency of each power generation unit;

[0092] The objective function is to maximize the output power of each generating unit without exceeding the load limit.

[0093] A constrained optimization algorithm is used to optimize the objective function to ensure that each power generation unit achieves maximum output power within its load constraint range while avoiding excessive load and power waste between units. The goal of the optimization algorithm is to minimize the following loss function:

[0094]

[0095] Where, P target，j is the target power of the jth power generation unit. Based on the output value of the photovoltaic power prediction model, the target power represents the theoretically optimal output power. The goal of the optimization algorithm is to make the actual output as close to this value as possible. λ is the weight coefficient between load regulation and safe operation. This coefficient determines the maximum load constraint P in the optimization process. max，j The importance of P max，j is the maximum load limit of the j-th power generation unit, indicating the maximum power value that the power generation unit can withstand. It means finding P in the optimization process unit The minimum value of .

[0096] The loss function is designed to find a balance between the target power and the maximum load, thereby ensuring that the power plant can operate efficiently in the short term while avoiding the risk of overloading the generating units.

[0097] In order to cope with complex weather conditions and sudden climate changes, the optimization algorithm will dynamically adjust the load distribution of the power generation unit. Under certain climatic conditions, the prediction error of photovoltaic power may be large. In this case, the target power P target，j The load adjustment weight λ allows for more flexible control of the output power of the power generation unit. For example, when the predicted power is lower than the actual generated power, the optimization algorithm appropriately increases the load to increase power output, while reducing the load to protect the equipment, ensuring the economic benefits of the power plant and the safety of the equipment.

[0098] Monitor real-time power generation data and dynamically adjust the dispatch plan if power generation deviates significantly from the forecast to prevent equipment overload or underutilization of power generation potential;

[0099] Real-time monitoring is achieved through the use of a variety of sensors installed in the power station, including temperature sensors, light sensors, humidity sensors, and current and voltage sensors. The data collected by the sensors is transmitted wirelessly to the central control system. The central control system analyzes the operating status and environmental changes of the power station in real time based on the sensor data. It can detect and dynamically adjust the errors between power prediction results and actual power generation, ensuring the accuracy and real-time performance of power optimization distribution.

[0100] Maximize the overall power generation efficiency of the power plant through optimized scheduling and real-time adjustments, while avoiding the risk of equipment damage and downtime caused by power scheduling errors;

[0101] The optimization algorithm for maximizing power generation efficiency further includes dynamically adjusting the scheduling strategy based on real-time weather changes in the power plant and the load conditions of the power generation units to cope with sudden weather changes. The optimization algorithm takes into account the impact of weather changes on photovoltaic power generation efficiency and improves power generation efficiency while ensuring equipment safety through a multi-objective optimization model, thereby enhancing the overall benefits of the photovoltaic power station.

[0102] Real-time adjustments are made by introducing edge computing devices, which can perform data processing and optimization calculations at the edge layer of the photovoltaic power station, reduce data transmission delays, and improve the system's response speed. The edge computing architecture can make power scheduling more flexible and responsive, especially when an emergency occurs in the power station. Edge computing can adjust power distribution in real time based on local data to ensure the continuous and efficient operation of the power station.

[0103] Implementation Method 1: To ensure efficient, optimized power allocation within a PV power plant in the short term, this implementation method introduces a machine learning algorithm to address the discrepancy between short-term PV power forecasts and actual power generation. Currently, PV power generation is affected by weather, equipment status, and environmental factors, making it difficult to avoid certain errors in power forecasts. This error can lead to "dynamic scheduling errors" during power dispatch within PV power plants, impacting their operational efficiency and safety. Therefore, accurate PV power forecasting and dynamic scheduling adjustments are key to optimizing PV power plant operations.

[0104] This implementation method first collects historical power generation data and real-time weather information (such as temperature, humidity, light intensity, wind speed, etc.) from photovoltaic power stations, pre-processes the data, removes abnormal data and noise, and ensures data reliability. Next, a power prediction model is established using deep learning algorithms in machine learning (such as convolutional neural networks (CNN) or long short-term memory networks (LSTM)). By learning information such as weather conditions and historical power generation data, the model can predict photovoltaic power output in the short term in the future. However, photovoltaic power generation has a strong time-varying nature, and the model may have certain errors during initial training, so the model needs to be continuously adjusted.

[0105] To address the problem of prediction errors, this implementation introduces a deviation analysis mechanism. Specifically, when there is a significant deviation between the actual power generated by a PV plant and the predicted value, the system performs an error analysis and calculates the mean absolute error (MAE) and root mean square error (RMSE). If the deviation exceeds a set threshold, the system dynamically adjusts the prediction model. A weighted average method is used to weight historical errors and real-time meteorological data, adjusting model parameters to ensure that the prediction results are more closely aligned with actual power generation conditions.

[0106] Furthermore, this implementation incorporates the operating status of power generation units into the power scheduling process. For example, if the system predicts that certain power generation units are about to become overloaded, it will prioritize reducing the output power of these units while increasing the power output of those with lighter loads. In this way, the PV power plant can avoid overloading localized equipment, thereby preventing equipment damage or power generation efficiency losses caused by overload operation.

[0107] When actual generated power falls below the predicted value, the system automatically adjusts the load on the generating units based on the forecast model's adjustments. This not only ensures the safe operation of the generating units but also maximizes power generation efficiency, avoiding missed power generation opportunities. Furthermore, the system dynamically adjusts power distribution among generating units in real time based on environmental changes and equipment conditions, maintaining efficient power plant operation.

[0108] By combining machine learning algorithms with a dynamic scheduling mechanism, this implementation can optimize the power distribution of photovoltaic power stations in the short term, not only improving power generation efficiency but also effectively avoiding equipment damage and system downtime caused by power prediction errors, ensuring the safe and stable operation of the power station.

[0109] Implementation Method 2: This implementation method uses edge computing technology to optimize the power scheduling and real-time adjustment of photovoltaic power plants, enabling more flexible and efficient power allocation in the face of complex weather conditions and equipment status changes. Traditional photovoltaic power plants typically rely on centralized data processing and scheduling strategies, which are subject to certain latency issues. Especially in the event of an emergency at the power plant, data transmission delays can lead to prolonged response times, impacting the plant's real-time scheduling capabilities. Therefore, the introduction of edge computing can effectively address this issue, improving the plant's responsiveness and flexibility.

[0110] First, edge computing devices locally process sensor data within the PV power plant. These sensors, including temperature, light, humidity, current, and voltage sensors, collect real-time operating status and environmental data from each power generation unit within the PV power plant. By transmitting this data to edge computing devices, the power plant can monitor the actual power generation status of each unit in real time, eliminating the need to wait for data to be transmitted to a central server for processing.

[0111] When a photovoltaic power plant encounters weather changes or equipment failures, edge computing devices can immediately process sensor data and respond quickly. For example, if sunlight intensity suddenly drops or cloud cover obscures sunlight, the plant's power generation capacity will fluctuate. In this case, edge computing devices adjust power distribution based on real-time environmental data and the load conditions of the power generation units to prevent overloading of individual units. By processing data locally, edge computing can quickly respond to these emergencies, enabling local power scheduling and ensuring that the overall power generation efficiency of the power plant is not affected by large fluctuations.

[0112] Furthermore, edge computing devices can optimize power scheduling through machine learning algorithms. Based on the operating status of each power generation unit within the power plant, the edge computing system can analyze in real time which units are lightly loaded and which are approaching overload, dynamically adjusting the power output of these units. The system also monitors the health of each unit. If the health of a unit deteriorates, the system reduces its load and dispatches other equipment to supplement its power generation, preventing equipment failure or unit downtime.

[0113] Through this localized optimization and scheduling enabled by edge computing, power plants can achieve more efficient power distribution, maximize power generation efficiency, and reduce the risks associated with delays and scheduling errors. Especially when power plants encounter emergencies, edge computing can significantly improve their response speed, ensuring that adjustments can be made in the shortest possible time, ensuring safe and stable operation.

[0114] Implementation Method 3: To further improve the accuracy of short-term PV power forecasts and effectively address dynamic scheduling issues caused by forecast errors, this implementation method introduces bias analysis and dynamic weighting adjustments within the power forecast model. This method improves the accuracy and stability of PV power plant power scheduling through real-time monitoring and analysis of forecast errors and algorithm optimization based on error adjustments, thereby addressing issues such as overloads, equipment failures, and inefficient power generation caused by forecast errors.

[0115] First, the implementation process begins by acquiring historical power generation data and real-time environmental data from the PV power plant. The system predicts the short-term future PV power generation based on meteorological factors at the PV plant's location (such as temperature, humidity, wind speed, cloud changes, etc.) and historical power generation records. This input data undergoes multi-dimensional data processing and feature extraction to form a high-quality dataset for model training. Machine learning models, particularly deep learning algorithms such as long-short-term memory (LSTM) networks, can identify the temporal patterns of PV power generation and predict PV power output over a period of time. However, because the PV power generation process is affected by multiple environmental variables, there is often a discrepancy between the predicted PV power and the actual power generation value.

[0116] To address this issue, the system performs real-time forecast error analysis. During power plant operation, the system continuously compares the actual generated power with the forecasted power at each moment, calculating the forecast error for each time period. If the forecast power for a particular time period deviates significantly from the actual power generation, the system immediately records and analyzes the deviation. By calculating the accumulated errors over the past few hours or even days, the system can identify time periods with the most severe forecast errors. Statistical analysis then assesses whether these errors are systematic deviations or fluctuations caused by sporadic environmental factors.

[0117] For predictions with large deviations, the system uses a dynamic weighted average method to adjust the model's prediction parameters. This adjustment process is based on the size of the error and the environmental characteristics at the time of prediction. Specifically, when the system finds that the prediction error for a certain period is large, it adjusts the model's learning parameters, especially the weight coefficients, based on the error situation for that period. This allows the model to pay more attention to periods with large errors and automatically adjusts the prediction results to reduce the recurrence of similar errors. For example, under extreme weather conditions, such as sudden cloud cover, the system adjusts the model's prediction behavior by increasing the weight of factors such as cloud cover and humidity in the meteorological data, thereby improving the prediction accuracy under similar meteorological conditions.

[0118] This dynamic weighted adjustment is not limited to adjusting the forecast error at a specific moment; it also encompasses the optimization of the entire PV power plant's power generation strategy. In actual operation, when a power plant encounters a large forecast error, the system will reallocate power based on the latest deviation analysis results. Specifically, the system dynamically adjusts the output power of each power generation unit to ensure that the overall power generation is as close to the forecast value as possible while avoiding overload or inefficient operation. If a power generation unit in a certain area exceeds a safe range due to excessive load, the system will immediately adjust the load in that area to restore it to a safer and more stable level, thereby avoiding damage or downtime to the inverter or other equipment.

[0119] At the same time, if other areas within the PV plant experience inefficient power generation, the system automatically dispatches power based on real-time power generation data, allowing these areas to maximize power generation without exceeding their load capacity. For example, certain generating units may perform poorly under certain weather conditions. The system will optimize scheduling strategies to allocate more power to other generating units to compensate for the shortfall in these areas, ensuring the overall power generation efficiency of the plant is not affected.

[0120] Another key feature of the system is its ability to respond quickly to changing weather conditions. Photovoltaic power generation is significantly affected by weather changes, especially when the weather fluctuates dramatically over a short period of time. Traditional fixed forecasting models often struggle to adapt to such changes. Therefore, during implementation, the system combines real-time changes in meteorological data with historical deviation analysis to rapidly adjust the forecasting model, enabling it to promptly adapt to new environmental changes. In inclement weather, the system increases the weighting of influencing factors such as cloud cover and wind speed, increasing sensitivity to meteorological changes and thus reducing forecast errors. This real-time adjustment capability ensures that the system can promptly and accurately optimize power allocation in response to sudden weather changes, minimizing losses.

[0121] Through this real-time dynamic adjustment and weighted correction mechanism, PV power plants can effectively improve the accuracy of power distribution without adding additional load, avoiding the overload and inefficient power generation common in traditional power forecasting. This approach not only improves the power generation capacity of power plants in complex environments, but also reduces the failure rate of power plant equipment through dynamic monitoring and automatic optimization, extending its service life and improving overall economic benefits.

[0122] In summary, after implementing this technical solution, the PV power plant's power generation system can adaptively adjust based on real-time forecast errors, thereby resolving the issue of irrational power scheduling caused by forecast errors and ensuring the PV plant achieves optimal power generation efficiency in all weather conditions. Furthermore, through intelligent dynamic scheduling, the plant's operation becomes more efficient and stable, while also avoiding issues such as equipment overload and downtime, significantly improving the safety and economic benefits of the PV plant.

[0123] This invention effectively improves the accuracy of short-term photovoltaic power forecasts. By introducing a machine learning-based photovoltaic power forecasting model and integrating it with historical power generation data and real-time weather information for refined processing, the present invention significantly improves the accuracy of power forecasts. Especially under complex meteorological conditions, by dynamically adjusting the model parameters, forecast errors are reduced, minimizing the deviation between the forecast results and actual power generation capacity, thereby providing a more accurate basis for subsequent power optimization and allocation. This method enables power plants to maintain efficient and stable power output under rapidly changing environmental conditions.

[0124] This invention effectively avoids power scheduling errors and improves power plant safety. In traditional photovoltaic power optimization and allocation processes, overscheduling or underloading due to prediction errors often leads to equipment damage or system downtime. This invention uses dynamic scheduling and real-time monitoring technology to ensure that the load of each power generation unit is controlled within a safe range, avoiding equipment overload. Through real-time data feedback and dynamic adjustments from intelligent sensors, power plants can quickly identify anomalies and take timely measures, effectively reducing the risk of equipment failure and enhancing the safety and reliability of power plants.

[0125] This invention significantly improves the overall power generation efficiency of photovoltaic power plants. Through an optimized power scheduling strategy, it maximizes the potential of photovoltaic power generation while avoiding equipment damage or insufficient power generation due to overscheduling. Especially in unstable weather conditions, the power plant can adjust the scheduling plan in real time to optimize load distribution based on actual power generation conditions, thereby ensuring efficient operation of the power generation units. Ultimately, the overall power generation efficiency of the power plant is significantly improved, increasing economic returns and reducing operating costs.

[0126] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0127] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0128] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0129] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0130] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0131] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0133] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0134] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0135] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power forecasting, characterized in that: The following steps are involved: Obtain historical power generation data and real-time weather information from photovoltaic power plants, remove noise data through data preprocessing, and generate input data sets for prediction; A photovoltaic power prediction model is established using a machine learning algorithm to predict photovoltaic power in the short term using the input data set; Analyze the deviation between the predicted power and the actual generated power, and adjust the parameters of the photovoltaic power prediction model according to the deviation; Based on the adjusted power forecast value and the operating status of each area in the power station, the power of the photovoltaic power generation unit is dispatched; Monitor real-time power generation data and dynamically adjust the dispatch plan if power generation deviates significantly from the forecast value; By optimizing scheduling and real-time adjustments, the overall power generation efficiency of the power station can be improved, and the risk of equipment damage and downtime caused by power scheduling errors can be avoided.

2. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: The machine learning algorithm further includes a feature engineering step, which extracts key features that affect photovoltaic power by analyzing historical power generation data and weather information of photovoltaic power stations. Through data standardization and normalization, the input data set has precise prediction accuracy.

3. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: The deviation analysis step further includes statistical analysis of the deviations of each power forecast result. The parameters of the photovoltaic power prediction model are adjusted by using a dynamic weighted average method. The weighted average method weights each deviation based on historical deviation values ​​and real-time weather changes, introducing higher adaptability into the prediction process.

4. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: The specific steps of deviation analysis are as follows: Obtain the actual power generation value P at each prediction moment from the real-time monitoring system of the photovoltaic power station actual And the predicted power value P forecast , where P actual is the actual generated power value obtained from the real-time measurement equipment, P forecast It is the predicted power value obtained by the photovoltaic power prediction model based on historical data and meteorological conditions; The prediction error at each moment is calculated based on the acquired data. The calculation expression is as follows: e t =P forecast，t -P actual，t Where, P actual，t is the actual power generation value at time t, P forecast，t is the predicted power value at time t, ε t is the forecast error at time t; By accumulating the prediction errors at all times, the mean absolute error is calculated as the overall performance indicator of the prediction model. The calculation expression is as follows: Where MAE is the mean absolute error, N is the total number of time points in the prediction period; Calculate the root mean square error to measure the accuracy of the prediction model. The root mean square error calculation expression is as follows: Where RMSE is the root mean square error; The overall prediction accuracy is calculated based on MAE and RMSE. The calculation expression is as follows: Where ρ is the overall prediction accuracy, and β is the adjustment factor used to adjust the impact of MAE and RMSE on the overall prediction accuracy.

5. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 4, characterized in that: The prediction model is adjusted based on the overall prediction accuracy ρ. The adjustment logic is as follows: If the overall accuracy ρ reaches the preset standard, it indicates that the prediction model performs well overall, and power scheduling can be directly performed and the optimization allocation process can be started; If the overall prediction accuracy ρ is lower than the preset standard, further local adjustments are performed.

6. The method for optimizing active power distribution of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: Based on the adjusted power forecast value and the operating status of each area in the power station, the power of the photovoltaic power generation unit is dispatched. The power dispatch steps are as follows: Define the load constraints for each photovoltaic power generation unit as follows: L min ≤P unit，j ≤L max Among them, L min is the minimum load limit of the j-th generating unit, L max is the maximum load limit of the jth generating unit, P unit，j is the actual output power of the jth power generation unit; Set the objective function F(P unit ) is the power generation benefit function of the entire power station. The goal is to maximize the output power of the power station, and considering the efficiency of each power generation unit, the objective function formula is as follows: Where W j is the efficiency weight of the jth power generation unit, M is the total number of power generation units in the photovoltaic power station, F(P unit ) is the overall power generation efficiency objective function of the PV power station, which represents the sum of the products of the output power of all power generation units and their efficiency weights; The constrained optimization algorithm is used to optimize the objective function to ensure that each power generation unit achieves the maximum output power within its load constraint range. The expression of the constrained optimization algorithm is: Where, P target，j is the target power of the jth power generation unit, based on the output value of the photovoltaic power prediction model, λ is the weight coefficient for load regulation and safe operation, P max，j is the maximum load limit of the j-th generating unit, It means finding P in the optimization process unit The minimum value of .

7. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: Real-time monitoring is achieved through the use of a variety of sensors installed in the power station, including temperature sensors, light sensors, humidity sensors, and current and voltage sensors. The data collected by the sensors is transmitted wirelessly to the central control system. The central control system analyzes the operating status and environmental changes of the power station in real time based on the sensor data, and realizes error detection and dynamic adjustment between power prediction results and actual power generation.

8. The method for optimizing active power distribution of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: The optimization algorithm for maximizing power generation efficiency further includes dynamically adjusting the scheduling strategy based on real-time weather changes in the power plant and the load conditions of the power generation units to cope with sudden weather changes. The optimization algorithm considers the impact of weather changes on photovoltaic power generation efficiency and improves power generation efficiency while ensuring equipment safety through a multi-objective optimization model.

9. The method for optimizing active power allocation of a photovoltaic power station based on short-term photovoltaic power prediction according to claim 1, characterized in that: Real-time adjustments are made by introducing edge computing devices, which can process data and optimize calculations at the edge layer of the photovoltaic power station, reducing data transmission delays and improving response speed; Through the edge computing architecture, power scheduling is made more flexible and responsive.

Images