Method and device for determining scheduling strategy of distributed photovoltaic power station and electronic equipment

Abstract

The application discloses a kind of distributed photovoltaic power station's scheduling strategy determination method, device and electronic equipment.Distributed photovoltaic power station's scheduling strategy determination method, device and electronic equipment are disclosed in the present application, and related technical field in new energy power generation scheduling is determined based on the historical meteorological data set of multiple photovoltaic power stations each in the region, determines the current operating condition corresponding to multiple photovoltaic power stations each;Optimize initial power generation output model to obtain target power generation output model;Using target power generation output model, obtain the predicted power generation output of target photovoltaic power station in current period;Based on the actual power generation output and predicted power generation output of target photovoltaic power station in current period, determine the current scheduling strategy of target photovoltaic power station.The present application solves the technical problems that the difference of meteorological conditions and meteorological observation station distribution between each distributed region of distributed photovoltaic power station is not considered enough when distributed photovoltaic power station is scheduled in complex operating environment in related art, resulting in low prediction accuracy of power generation output of distributed photovoltaic power station and scheduling response lag.

Description

Technical Field

[0001] This invention relates to the field of new energy power generation dispatching technology, and more specifically, to a method, apparatus, and electronic equipment for determining the dispatching strategy of a distributed photovoltaic power station. Background Technology

[0002] In related technologies, the dispatch strategies for distributed photovoltaic (PV) power plants often rely on observation data from single or limited meteorological stations, failing to fully consider the heterogeneity of meteorological conditions across different regions and the uneven distribution density of observation stations. Since distributed PV power plants typically cover vast geographical areas, meteorological parameters (such as irradiance, temperature, and wind speed) exhibit significant spatial heterogeneity due to factors like topography, altitude, and microclimate. Methods in related technologies often neglect the nonlinear impact of local meteorological characteristics on power generation output, resulting in poor generalization ability of prediction models. Furthermore, dispatch systems in related technologies are mostly based on static or general power generation models, lacking dynamic adaptation to different power plant operating conditions (such as equipment aging, tilt angle differences, and shading conditions), making it difficult to respond to sudden meteorological changes. This leads to the accumulation of output prediction errors and delayed dispatch response. Especially in scenarios involving the coordinated operation of multi-regional and multi-type distributed PV clusters, this deficiency further exacerbates grid peak-shaving pressure and the risk of curtailment, severely impacting the efficiency of new energy absorption and system stability. In summary, the relevant technologies for scheduling distributed photovoltaic power stations under complex operating environments do not adequately consider the differences in meteorological conditions and meteorological observation station distribution among the various distribution areas of distributed photovoltaic power stations, resulting in problems such as low accuracy of power generation prediction and delayed scheduling response.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This invention provides a method, apparatus, and electronic device for determining the scheduling strategy of a distributed photovoltaic power station, which at least solves the technical problems in related technologies where insufficient consideration is given to the differences in meteorological conditions and meteorological observation station distribution among the various distribution areas of the distributed photovoltaic power station when scheduling the distributed photovoltaic power station in a complex operating environment, resulting in low power generation prediction accuracy and delayed scheduling response of the distributed photovoltaic power station.

[0005] According to one aspect of the present invention, a method for determining the scheduling strategy of a distributed photovoltaic (PV) power station is provided, comprising: determining the current operating conditions of each of the multiple PV power stations based on historical meteorological datasets of the respective regions of the multiple PV power stations; determining an initial power generation output model based on historical power generation datasets of the multiple PV power stations and historical meteorological datasets of the respective regions of the multiple PV power stations, wherein the initial power generation output model is used to predict the power generation output of the PV power stations under different meteorological conditions; optimizing the initial power generation output model based on the current operating conditions and historical power generation datasets of the multiple PV power stations to obtain a target power generation output model; obtaining a predicted power generation output of the target PV power station in the current period based on the target power generation dataset of the target PV power station in the current time period and the target meteorological dataset of the region where the target PV power station is located in the current time period, using the target power generation output model; and determining the current scheduling strategy of the target PV power station based on the actual power generation output and predicted power generation output of the target PV power station in the current period.

[0006] According to another aspect of the present invention, a device for determining the scheduling strategy of a distributed photovoltaic power station is also provided, comprising: a current operating condition determination module, configured to determine the current operating conditions of each of the multiple photovoltaic power stations based on historical meteorological datasets of the respective regions of the multiple photovoltaic power stations; an initial power generation output model construction module, configured to determine an initial power generation output model based on historical power generation datasets corresponding to each of the multiple photovoltaic power stations and historical meteorological datasets of the respective regions of the multiple photovoltaic power stations, wherein the initial power generation output model is used to predict the power generation output of the photovoltaic power station under different meteorological conditions; a target power generation output model determination module, configured to optimize the initial power generation output model based on the current operating conditions and historical power generation datasets of each of the multiple photovoltaic power stations to obtain a target power generation output model; a predicted power generation output determination module, configured to obtain the predicted power generation output of the target photovoltaic power station in the current period based on the target power generation dataset of the target photovoltaic power station in the current period and the target meteorological dataset of the region where the target photovoltaic power station is located in the current period, using the target power generation output model; and a current scheduling strategy determination module, configured to determine the current scheduling strategy of the target photovoltaic power station based on the actual power generation output and predicted power generation output of the target photovoltaic power station in the current period.

[0007] According to another aspect of the present invention, a non-volatile storage medium is also provided, the non-volatile storage medium storing a plurality of instructions, the instructions being adapted to be loaded by a processor and executed by any one of the methods for determining the scheduling strategy of a distributed photovoltaic power station.

[0008] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method for determining the scheduling strategy of a distributed photovoltaic power station as described in any one of the present inventions.

[0009] According to another aspect of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the method for determining the scheduling strategy of a distributed photovoltaic power station as described in any one of the present invention.

[0010] In this embodiment of the invention, the current operating conditions of each photovoltaic power station are determined based on historical meteorological datasets of their respective regions. An initial power generation model is determined based on historical power generation datasets of each photovoltaic power station and historical meteorological datasets of their respective regions. This initial power generation model is used to predict the power generation output of the photovoltaic power stations under different meteorological conditions. The initial power generation model is optimized based on the current operating conditions and historical power generation datasets of each photovoltaic power station to obtain a target power generation model. Based on the target power generation dataset of the target photovoltaic power station in the current time period and the target meteorological dataset of the region where the target photovoltaic power station is located in the current time period, the target power generation model is used to obtain the predicted power generation output of the target photovoltaic power station in the current time period. Finally, based on the target photovoltaic power station's current power generation data, the predicted power generation output of the target photovoltaic power station in the current time period is obtained. This method uses the actual and predicted power output of a segment to determine the current dispatch strategy for the target photovoltaic power station. It achieves the goal of determining and optimizing the initial power output model using historical meteorological datasets from multiple regions and corresponding historical power generation datasets. The obtained target power output model is then used to determine the predicted power output of the target photovoltaic power station in the current time period. Based on the actual power output of the target photovoltaic power station in the current time period, the current dispatch strategy is accurately determined. This improves the accuracy of power output prediction and dispatch response efficiency of distributed photovoltaic power stations. Furthermore, it solves the technical problem in related technologies where insufficient consideration is given to the differences in meteorological conditions and meteorological station distribution among different areas of distributed photovoltaic power stations under complex operating environments, leading to low power output prediction accuracy and delayed dispatch response. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0012] Figure 1This is a flowchart of a method for determining the scheduling strategy of a distributed photovoltaic power station according to an embodiment of the present invention;

[0013] Figure 2 This is a flowchart of an optional method for determining a dimensionality reduction dataset according to an embodiment of the present invention;

[0014] Figure 3 This is a flowchart of an optional method for determining the scheduling strategy of a distributed photovoltaic power station according to an embodiment of the present invention;

[0015] Figure 4 This is a schematic diagram of an optional distributed photovoltaic power station scheduling strategy determination system according to an embodiment of the present invention;

[0016] Figure 5 This is a schematic diagram of a scheduling strategy determination device for a distributed photovoltaic power station according to an embodiment of the present invention. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] According to an embodiment of the present invention, a method embodiment for determining the scheduling strategy of a distributed photovoltaic power station is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0020] Figure 1This is a flowchart of a method for determining the scheduling strategy of a distributed photovoltaic power station according to an embodiment of the present invention, as follows: Figure 1 As shown, the method includes the following steps:

[0021] Step S102: Based on the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located, determine the current operating conditions of each of the multiple photovoltaic power stations.

[0022] Optionally, historical meteorological datasets are extracted from meteorological observation stations in the respective regions of multiple photovoltaic power plants. These datasets, combined with the geographical characteristics of each region (such as latitude, altitude, terrain type, and climate zone), are analyzed to understand the dynamic impact of different meteorological data (such as irradiance, temperature, wind speed, humidity, and cloud cover) on power generation. The historical meteorological datasets include meteorological data collected over predetermined historical periods. By analyzing these datasets from multiple regions, executable current operating conditions are generated for each photovoltaic power plant. These current operating conditions include, but are not limited to: seasonal upper limit thresholds for power generation, temperature-sensitive derating strategies, cleaning and maintenance windows triggered by meteorological cycles, and irradiance response sensitivity coefficients. The current operating conditions are a set of predictive and strategic operating rules derived from the correlation between historically accumulated meteorological data and power generation, rather than instantaneous meteorological data. This step, constructing current operating conditions from meteorological datasets, transforms historical experience into rules, enabling photovoltaic power plants to shift from passive response to proactive adaptation, achieving a leap from data-driven prediction to strategy-driven operation.

[0023] In an optional embodiment, before determining the current operating conditions of each of the multiple photovoltaic power stations based on historical meteorological datasets of their respective regions, the method further includes: determining the meteorological observation range of any photovoltaic power station's region based on the location information of a meteorological observation station in the region where any photovoltaic power station is located, wherein the location information includes at least: latitude and longitude, altitude, and the geographical environment of the region where the meteorological observation station is located; conducting meteorological observations on the region where any photovoltaic power station is located according to the meteorological observation range based on a preset meteorological observation cycle to determine the initial meteorological dataset of the region where any photovoltaic power station is located; optimizing the initial meteorological dataset to obtain the historical meteorological dataset of the region where any photovoltaic power station is located; and obtaining the historical meteorological datasets of the regions where multiple photovoltaic power stations are located by using the method of obtaining the historical meteorological dataset of the region where any photovoltaic power station is located.

[0024] Optionally, before determining the current operating conditions of multiple photovoltaic power stations, a refined meteorological data collection method can ensure that the historical meteorological dataset is observational data that has been precisely matched to the geographical environment of each photovoltaic power station, possesses spatial representativeness, and has undergone quality verification. Specifically, firstly, for any photovoltaic power station's location, the precise location information of surrounding meteorological observation stations is obtained. This location information includes, but is not limited to, latitude and longitude, altitude, and the geographical environmental characteristics of the observation station (such as terrain undulations, surrounding obstructions), and land cover type (desert, vegetation, water surface). This location information determines the effective sensing space of the meteorological observation station, that is, the coverage area of ​​meteorological features that its sensors can truly reflect. For example, the wind speed and temperature data of an observation station located at the bottom of a valley cannot represent the environment of a photovoltaic power station on a mountaintop; the humidity data of an observation station near a lake may be high, but the irradiance measurement may be distorted due to water vapor evaporation. Therefore, based on the technical parameters of meteorological observation stations (such as radar detection radius and sensor effective distance) and the geographical environment, a three-dimensional meteorological observation range for each meteorological observation station is constructed and spatially overlaid with a regional boundary map to achieve accurate spatial mapping between the meteorological observation station and the photovoltaic power station area. On this basis, meteorological data is continuously collected for each area according to a preset meteorological observation cycle (e.g., data collected every 10 minutes, calibrated daily) to form an initial meteorological dataset. However, the meteorological data in the initial dataset may contain noise, missing values, or equipment drift. Therefore, further optimization of the initial meteorological dataset is necessary. This optimization can be achieved, but is not limited to, through the following methods: regularly testing sensor performance based on the meteorological observation station's operation and maintenance records and equipment calibration cycle; automatically identifying and correcting outliers (e.g., a sudden temperature increase of 50°C), data breaks (e.g., no transmission for 30 consecutive minutes), or systematic deviations (e.g., irradiance 15% lower); or data completion and correction through interpolation from neighboring stations and historical trend regression, ultimately generating a high-quality, reliable historical meteorological dataset. The historical meteorological dataset collection process is not repeated for a single photovoltaic power station, but is carried out in parallel and synchronously for all photovoltaic power station areas, based on regions, observation stations, and photovoltaic power stations. This ensures that each photovoltaic power station has a historical meteorological dataset that is highly consistent with its geographical environment and of reliable quality, so that the operating conditions of each photovoltaic power station are based on data that truly belong to its local climate background.

[0025] In one optional embodiment, based on historical meteorological datasets of the respective regions of multiple photovoltaic power stations, the current operating conditions of each photovoltaic power station are determined, including: determining multiple power generation impact values ​​for any photovoltaic power station based on historical meteorological datasets of the region where any photovoltaic power station is located, wherein the multiple power generation impact values ​​are used to quantify the influence intensity of multiple historical meteorological data in the historical meteorological dataset of the region where any photovoltaic power station is located on the power generation output of any photovoltaic power station, and the multiple power generation impact values ​​correspond one-to-one with the multiple historical meteorological data; determining the power generation impact value that exceeds a preset impact threshold among the multiple power generation impact values ​​as the power generation impact factor of any photovoltaic power station; determining the current operating conditions of any photovoltaic power station based on the power generation impact factor; or, if none of the multiple power generation impact values ​​exceed the preset impact threshold, determining the current operating conditions of any photovoltaic power station based on the historical operating conditions of any photovoltaic power station; obtaining the current operating conditions of each of the multiple photovoltaic power stations by using the method of obtaining the current operating conditions of any photovoltaic power station.

[0026] Optionally, firstly, for each meteorological data point (such as irradiance, ambient temperature, wind speed, humidity, and cloud cover) in the historical meteorological dataset of any photovoltaic power station's location, analyze the impact strength of each meteorological data point on power generation output. This yields a power generation impact value for any photovoltaic power station. This impact value is not a static correlation coefficient, but rather calculated by constructing a difference model between meteorological changes and power generation response. The relative rate of change in power generation per unit change in meteorological data within a specific interval is calculated, and a weighted average is applied to a large number of historical samples to ultimately derive the impact strength index for each meteorological data point. Subsequently, the power generation impact value is compared with a preset impact threshold, which can be comprehensively set based on the photovoltaic power station's design capacity, historical operation and maintenance experience, and grid safety boundaries. Only when the impact strength of a certain meteorological data point exceeds this threshold is it considered a power generation influencing factor. Furthermore, by considering the direction and trend of the power generation influencing factors, the current operating conditions for any photovoltaic power station are constructed. For example, in a foggy coastal area, humidity has a significant impact while temperature has a weak impact; therefore, the current operating conditions could be to initiate dehumidification and delayed cleaning on foggy days. Conversely, if none of the multiple power generation impact values ​​of any photovoltaic power station exceed the preset impact threshold, then the system no longer relies on weather-driven factors. Instead, it is based on the historical operating conditions accumulated over a long period of time by the photovoltaic power station, such as equipment rated limits, annual maintenance plans, and past dispatch instructions. This ensures that the photovoltaic power station maintains a stable and predictable operating state in the absence of significant weather disturbances, avoiding excessive intervention for the sake of optimization. The method of determining the current operating conditions through this step can provide accurate and reliable control boundaries for subsequent power generation output model optimization, power generation output prediction, and dispatch response.

[0027] Step S104: Based on the historical power generation datasets corresponding to each of the multiple photovoltaic power plants and the historical meteorological datasets of the regions where the multiple photovoltaic power plants are located, determine the initial power generation output model, wherein the initial power generation output model is used to predict the power generation output of the photovoltaic power plants under different meteorological conditions.

[0028] Optionally, the power generation dataset (including but not limited to DC power, AC power, conversion efficiency, and operating temperature) of each photovoltaic power station is precisely matched temporally and spatially with the historical meteorological dataset of its region. This means the historical power generation dataset includes power generation data collected over a predetermined historical period. Based on this, an initial power generation output model is constructed using the historical power generation datasets corresponding to multiple photovoltaic power stations and the historical meteorological datasets of their respective regions. This initial power generation output model is independently modeled on a per-unit basis, incorporating the response patterns of the geographical environment in which the photovoltaic power station is located. For example, for photovoltaic power stations located in high-altitude arid areas, the initial power generation output model will learn the nonlinear law of significant efficiency improvement under the combination of high irradiance, low humidity, and strong winds; while for power stations located in low-lying, humid and hot areas, it will learn the unique behavior of high humidity and high temperature leading to aggravated hot spot effects and more significant power output decay. This initial power generation output model is an empirical model learned from real-world operation, capable of capturing the causal relationship between meteorological and power generation data. Ultimately, this initial power generation output model can output reasonable, reliable, and engineering-reference-worthy power generation output predictions for photovoltaic power stations under any meteorological conditions and in any region, without human intervention.

[0029] In one optional embodiment, an initial power generation model is determined based on the historical power generation datasets corresponding to each of the multiple photovoltaic power plants and the historical meteorological datasets of the regions where the multiple photovoltaic power plants are located. This includes: determining multiple historical correlation results based on the historical power generation datasets corresponding to each of the multiple photovoltaic power plants and the historical meteorological datasets of the regions where the multiple photovoltaic power plants are located, wherein the multiple historical correlation results correspond one-to-one with the multiple photovoltaic power plants, and the multiple historical correlation results are used to quantify the degree of correlation between the historical power generation datasets of the corresponding photovoltaic power plants and the historical meteorological datasets of the corresponding regions; optimizing and training the power generation network based on the historical power generation datasets corresponding to each of the multiple photovoltaic power plants, the historical meteorological datasets of the regions where the multiple photovoltaic power plants are located, the multiple historical correlation results, and the actual power generation output of each of the multiple photovoltaic power plants during the historical prediction period, with the optimization objective of minimizing the difference between any actual power generation output and the corresponding predicted power generation output, until a predetermined number of iterations is reached, wherein the corresponding predicted historical power generation output is obtained by using the power generation network to predict based on the historical power generation datasets, historical meteorological datasets, and historical correlation results corresponding to any of the multiple photovoltaic power plants during the optimization and training process; and obtaining the initial power generation model based on the optimized power generation network obtained after reaching the predetermined number of iterations.

[0030] Optionally, in this embodiment, instead of directly using historical power generation datasets and corresponding regional historical meteorological datasets to construct the initial power generation model, historical correlation results are introduced to quantify the correlation strength between the historical power generation datasets of each photovoltaic power station and the corresponding historical meteorological data. The power generation network is input with the historical power generation datasets of each photovoltaic power station, the historical meteorological datasets of the corresponding region, and the corresponding historical correlation results. The output is the predicted power generation output of the photovoltaic power station in the historical prediction period. The historical prediction period is the period following the predetermined historical period; the predetermined historical period is the period during which historical power generation datasets corresponding to multiple photovoltaic power stations and historical meteorological datasets of the regions where the multiple photovoltaic power stations are located are collected. During training, the power generation network does not treat all photovoltaic power station data as a single entity. Instead, in each training round, its internal parameter weights are dynamically adjusted based on the historical correlation results of each photovoltaic power station, allowing the power generation network to remember the differentiated operating logic of different regions. For example, when the power generation network learns that the temperature decay effect of the first photovoltaic power station is significant, while the wind speed gain of the second photovoltaic power station is significant, the power generation network will automatically enhance the neuron response of the temperature channel when processing the data of the first photovoltaic power station, and activate the feature mapping path related to wind speed when processing the data of the second photovoltaic power station. This training mechanism enables the power generation network to autonomously learn and solidify the characteristics of each photovoltaic power station while sharing a unified architecture. This training process continues until a preset number of iterations is reached. At this point, the power generation network no longer fluctuates drastically due to data disturbances, and its adaptability to regional differences has stabilized sufficiently. This power generation network is then solidified as the initial power generation model. This initial model is not a model specific to a particular region, nor is it an average model that ignores differences. Instead, it is a model capable of identifying which photovoltaic power station region any meteorological dataset belongs to, automatically recalling historical correlation results, and accurately outputting predicted power generation output.

[0031] Optionally, the power generation network can adopt a dual-channel heterogeneous deep architecture, consisting of a regional feature encoder, a power plant-specific mapper, and a fusion predictor. The overall structure is an end-to-end supervised deep regression network, with the optimization objective being to minimize the mean square error between the predicted power generation output and the corresponding actual power generation output for historical prediction periods. Iterative optimization is performed through backpropagation. Specifically, firstly, the regional feature encoder processes the historical meteorological dataset associated with each photovoltaic power plant. The historical meteorological dataset of the photovoltaic power plant region is concatenated with the region's geographical features to form a high-dimensional spatiotemporal feature vector. This vector is input into a lightweight hybrid encoder: using a preset sliding window, it extracts local patterns from the historical meteorological dataset (such as trends in sudden drops in irradiance and periodic temperature fluctuations), while simultaneously capturing long-term dependencies across days and seasons (such as the cumulative effect of summer high temperatures and the delayed response to winter morning fog). The encoder outputs a regional-level feature vector. Next, the historical association results are directly injected into the power generation network as regional-specific weights prior. The historical association results serve as an attention gating factor, dynamically weighting the contribution of each meteorological channel during meteorological data extraction. For example, if the effect of temperature on a certain region has a preset critical value, the power generation network automatically suppresses gradient backpropagation in the temperature channel, ignoring the interference of this temperature-related variable. Then, the power plant-specific mapper receives two inputs: a regional feature vector output by the regional encoder and the historical power generation dataset of the photovoltaic power plant itself. This standardized historical power generation dataset is input to the encoder to extract the photovoltaic power plant's behavioral patterns (such as whether there is degradation, whether efficiency decreases due to component tilt aging, whether there are fixed-period fluctuations due to shading, etc.). The encoder outputs a power plant-specific feature vector. Next, the regional feature vector and the power plant-specific feature vector are fed into a fully connected residual network. This fully connected residual network can consist of multiple hidden layers, each containing a predetermined number of neurons, and introduces skip connections to alleviate gradient vanishing, ultimately outputting the joint features of the power plant and the region. Finally, the fusion predictor inputs the joint features of the power plant and the region into a two-layer feedforward network, and the output is the predicted power generation for the historical prediction period. Throughout the entire power generation network training process, the actual historical power output is used as the label, and an optimizer is employed, with the loss function being the mean squared error. The training process continues until the preset number of iterations is reached. At this point, the parameters of the power generation network converge, which is the final initial power generation model.

[0032] Step S106: Based on the current operating conditions and historical power generation datasets of each of the multiple photovoltaic power plants, optimize the initial power generation output model to obtain the target power generation output model.

[0033] Optionally, while the initial power generation model possesses regionally differentiated prediction capabilities, its training relies on statistical patterns from historical data and does not yet incorporate the current operating conditions of the photovoltaic power plant. For example, a photovoltaic power plant may be mandated to prioritize power consumption during low-load periods due to grid safety requirements. These constraints are not determined by weather but are imposed by operation and maintenance strategies, safety procedures, or dispatch instructions. If the initial power generation model continues to predict power generation based on historical data, biases will arise, leading to inaccurate predictions. Therefore, the optimization process should consider current operating conditions and use historical power generation data as a foundation to guide the parameter adjustments of the initial power generation model. The resulting target power generation model can more accurately predict power generation, thereby improving the precision of the dispatch strategy.

[0034] In one optional embodiment, an initial power generation model is optimized based on the current operating conditions and historical power generation datasets of multiple photovoltaic power plants to obtain a target power generation model. This includes: obtaining a power generation feature set based on the current operating conditions and historical power generation datasets of the multiple photovoltaic power plants, wherein the power generation feature set is a collection of features associated with the power generation output of the multiple photovoltaic power plants; performing feature transformation on the power generation feature set to obtain a feature transformation dataset, wherein the feature transformation converts the power generation feature set into a dataset that conforms to the input dimension of the initial power generation model; if the number of features in the feature transformation dataset exceeds a preset number, performing dimensionality reduction on the feature transformation dataset to obtain a dimensionality-reduced dataset; optimizing the initial power generation model based on the dimensionality-reduced dataset to obtain the target power generation model; or optimizing the initial power generation model based on the feature transformation dataset if the number of features in the feature transformation dataset does not exceed a preset number to obtain the target power generation model.

[0035] Optionally, firstly, to construct a power generation output feature set, spatiotemporal alignment and semantic fusion are performed on the current operating conditions and historical power generation datasets corresponding to multiple photovoltaic power plants. All historical power generation data associated with the power generation output under the current operating conditions are extracted to form a power generation output feature set. This feature set includes, but is not limited to, irradiance, temperature, equipment status, and operating mode, thus comprehensively representing how much electricity the photovoltaic power plant can actually generate under the current operating conditions. Subsequently, feature transformation is performed on the feature transformation dataset. This involves operations such as missing value imputation, outlier smoothing, non-numerical variable encoding, and time window aggregation to uniformly transform the feature transformation dataset into a structurally regular, numerically continuous feature transformation dataset that conforms to the input specifications of the initial power generation model. Further, when the number of features in the transformed feature transformation dataset exceeds a preset threshold, a linear discriminant analysis dimensionality reduction mechanism is activated. While preserving the original feature information, this mechanism actively seeks a low-dimensional projection space that can best distinguish different operating states, compressing high-dimensional features to low-dimensional features below the preset threshold. This dimensionality reduction process can significantly reduce the input dimensionality of the initial power generation model and reduce the computational burden. Conversely, if the number of features does not exceed a preset limit, the feature-transformed dataset is used directly for optimization to avoid unnecessary dimensionality reduction losses. Ultimately, regardless of whether dimensionality reduction is performed, the feature-transformed dataset is used as input to optimize the parameters of the initial power generation model. During optimization, the focus is on learning the optimal mapping relationship between the current operating conditions and historical power generation datasets, thereby eliminating noise interference, removing redundant correlations, and converging to the most discriminative power generation prediction path.

[0036] In one alternative embodiment, Figure 2 This is a flowchart of an optional method for determining a dimensionality reduction dataset according to an embodiment of the present invention, such as... Figure 2 As shown, when the number of features in the feature transformation dataset exceeds a preset number, the feature transformation dataset is dimensionality reduced to obtain a dimensionality-reduced dataset, including:

[0037] Step S202: When the number of features in the feature transformation dataset exceeds a preset number, the feature data in the feature transformation dataset is classified based on multiple preset category labels to obtain multiple classification subsets. The multiple preset category labels include at least a high output status label, a medium output status label, and a low output status label. The high output status label is used to identify feature data whose power generation output is greater than or equal to a preset first threshold. The medium output status label is used to identify feature data whose power generation output is between the preset first threshold and a preset second threshold. The low output status label is used to identify feature data whose power generation output is less than or equal to a preset second threshold. The preset first threshold is greater than the preset second threshold. The multiple classification subsets each include feature data corresponding to the preset category labels.

[0038] Step S204: Determine the mean vector and covariance matrix corresponding to each of the multiple classification subsets. Each element in the mean vector is the mean of the samples included in the corresponding feature data of the corresponding classification subset. Each off-diagonal element in the covariance matrix is ​​the covariance between the samples included in the two feature data of the corresponding classification subset, and each diagonal element is the variance of the samples included in the corresponding feature data of the corresponding classification subset.

[0039] Step S206: Based on the mean vectors corresponding to each of the multiple classification subsets, the inter-class scatter matrix is ​​obtained, wherein the inter-class scatter matrix is ​​used to quantify the degree of dispersion among the multiple classification subsets;

[0040] Step S208: Based on the covariance matrix corresponding to each of the multiple classification subsets and the total number of samples corresponding to each of the multiple classification subsets, the intra-class scatter matrix is ​​obtained. The intra-class scatter matrix is ​​used to quantify the degree of dispersion among multiple feature data in any one of the multiple classification subsets.

[0041] Step S210: Perform generalized eigenvalue decomposition on the inter-class scatter matrix and the intra-class scatter matrix to obtain the target generalized eigenvector, wherein the target generalized eigenvector represents the generalized eigenvector corresponding to the generalized eigenvalue that is greater than the preset eigenvalue threshold among multiple generalized eigenvalues.

[0042] Step S212: When there are multiple target generalized feature vectors, the multiple target generalized feature vectors are juxtaposed in the column direction based on a preset sorting method to obtain the feature vector projection matrix. The preset sorting method is the sorting of the generalized feature values ​​corresponding to the multiple target generalized feature vectors from largest to smallest.

[0043] Step S214: Based on the feature vector projection matrix, perform dimensionality reduction processing on the feature transformation dataset to obtain a dimensionality-reduced dataset.

[0044] Optionally, when the number of features in the feature transformation dataset exceeds a preset threshold, dimensionality reduction is performed using linear discriminant analysis guided by category labels. Essentially, this compresses redundant feature dimensions while preserving maximum discriminative information, allowing the initial power generation output model to focus more on key variables that are discriminative of power generation output status. Specifically, firstly, based on multiple preset category labels, a preset category label is assigned to each feature data in the feature transformation dataset, ultimately determining the subset to which the feature data belongs. Multiple subsets are obtained by using the method of obtaining the subset to which any feature data belongs. Next, the mean vector and covariance matrix corresponding to each subset are determined. For example, taking any feature data (current) in the subset corresponding to high output status as an example, this current includes current values ​​(samples) collected at multiple different historical times. Any element in the mean vector of any subset can be obtained as the average value A of the samples included in the current, which is: ,in, Let represent the current value collected at any given historical moment, where 'i' represents the index of that historical moment, and 'M' represents the number of historical moments (samples). For example, taking two feature data points (current and irradiance) in a subset belonging to the high-output state as an example, where current and irradiance each contain multiple current and irradiance values ​​(samples) collected at different historical moments, any off-diagonal element Cov(Current, Irradiance) in the covariance matrix of any subset can be obtained as follows: ,in, This represents the irradiance value collected at any given historical moment. This indicates that the irradiance includes the average value of the sample. The value is obtained by calculating the average of the samples included in the current. Taking any feature data (current) in the subset corresponding to the high-output state as an example, any diagonal element Var(Current) in the covariance matrix of any subset can be obtained as follows: By obtaining any element from the mean vector of any subset of classifications, and any off-diagonal and diagonal elements from the covariance matrix of any subset of classifications, we can obtain the mean vectors and covariance matrices corresponding to multiple subsets of classifications. Subsequently, based on the mean vectors of each subset of classifications, we can calculate the inter-class scatter matrix. Any off-diagonal element in the inter-class scatter matrix can be obtained as follows: ,in, This represents the average value of the samples included in any one of the features within a subset of multiple classification data. Let C represent the average value of samples in multiple subsets that include data from multiple features, excluding any one of the features. Let C represent the number of subsets. Let j represent the average value of the samples included in any feature data within any subset of categories, j represent any feature data within any subset of categories, and k represent another feature data within any subset of categories besides any other feature data. It represents the average value of the samples included in any subset of categories, which is another feature data besides any other feature data. The scatter matrix represents the total number of samples included in multiple feature data within any given subset, i.e., the total number of samples in any given subset. N represents the total number of samples included in multiple feature data across multiple subsets, and P represents the index of any off-diagonal element. Any diagonal element in the inter-class scatter matrix can be obtained as follows: Where W represents the index of any diagonal element. Further, based on the covariance matrices of each of the multiple classification subsets and the total number of samples corresponding to each of the multiple classification subsets, the within-class scatter matrix is ​​obtained. Any off-diagonal element in the within-class scatter matrix can be obtained as follows: ,in, This represents the covariance between any feature data in any subset of categories and another feature data besides that feature data. This represents any sample value of any feature data in any subset of categories. Let represent any sample value of another feature data in any subset of classifications, excluding any other feature data, and let i represent the index of any sample of any feature data in any subset of classifications. Any diagonal element in the intra-class scatter matrix can be obtained as follows: , Next, generalized eigenvalue decomposition is performed on the inter-class and intra-class scatter matrices to obtain a set of generalized eigenvalues ​​and their corresponding generalized eigenvectors. The larger the generalized eigenvalue, the more significant the inter-state difference relative to intra-state fluctuations along that eigenvector direction. Only eigenvectors with generalized eigenvalues ​​exceeding a preset threshold are retained as target generalized eigenvectors, ensuring that only truly discriminative directions are preserved after dimensionality reduction. When there are multiple target generalized eigenvectors, they are further sorted in descending order of their corresponding generalized eigenvalues, and these eigenvectors are concatenated as columns to form an eigenvector projection matrix. Finally, each feature data in the feature transformation dataset is multiplied by this eigenvector projection matrix to obtain a lower-dimensionality dataset, retaining the most discriminative power generation output feature combinations, thereby significantly improving the accuracy of subsequent target power generation output model predictions.

[0045] Step S108: Based on the target power generation dataset of the target photovoltaic power station in the current time period among multiple photovoltaic power stations, and the target meteorological dataset of the area where the target photovoltaic power station is located in the current time period, the target power generation output model is used to obtain the predicted power generation output of the target photovoltaic power station in the current time period.

[0046] Optionally, by acquiring the target power generation dataset of the target photovoltaic power station in the current time period and the target meteorological dataset of the area where the target photovoltaic power station is located in the current time period, and using a pre-trained target power generation output model, the predicted power generation output of the target photovoltaic power station in the current time period can be obtained, thereby enabling real-time prediction of photovoltaic power generation output. The target meteorological dataset includes meteorological data collected in the current time period, and the target power generation dataset includes power generation data collected in the current time period.

[0047] Step S110: Determine the current dispatch strategy for the target photovoltaic power station based on the actual and predicted power output of the target photovoltaic power station in the current time period.

[0048] Optionally, after obtaining the predicted power output of the target photovoltaic power station for the current period, the actual power output of the target photovoltaic power station for the current period can be collected simultaneously through on-site power measurement equipment (such as smart meters, data acquisition and monitoring control systems). Comparing the two can assess the reliability of the target power output model and the degree of abnormality in the operating status of the target photovoltaic power station, thereby formulating a dispatch strategy. This step can significantly improve the practicality of power output prediction and the proactivity of dispatch decisions. It can not only effectively suppress the impact of prediction deviations caused by meteorological data noise, equipment aging, or regional characteristic differences on grid stability, but also provide real-time and reliable guarantees for the safe and flexible operation of the new energy grid through dispatch strategies.

[0049] In one optional embodiment, the current scheduling strategy for the target photovoltaic power station is determined based on its actual and predicted power generation output during the current time period. This includes: determining the power generation error value of the target photovoltaic power station based on the actual and predicted power generation output; if the power generation error value exceeds a preset error value, determining the current scheduling strategy as compensating for the target photovoltaic power station, wherein the compensating scheduling includes at least one of the following: controlling the energy storage device to perform charging and discharging operations to compensate for the power generation error value, adjusting the reactive power output of the target photovoltaic power station to compensate for the power generation error value; or if the power generation error value does not exceed the preset error value, determining the current scheduling strategy as maintaining the current operating state of the target photovoltaic power station.

[0050] Optionally, the actual power output of the target photovoltaic power station, obtained from power measurement equipment during the current period, is compared in real time with the predicted power output derived from the target power output model, and the error value between the two is calculated. This error value not only directly reflects the accuracy of the target power output model but also carries implicit information about potential operational risks such as abnormal meteorological data, equipment status drift, or insufficient modeling of regional characteristics. When the error value exceeds a preset error value, it is determined that the prediction of the current target power output model deviates significantly from the actual operating state, exceeding the acceptable tolerance range. A compensation scheduling mechanism is then activated, which includes, but is not limited to: coordinating the charging and discharging operations of energy storage devices deployed in the power station or region to offset the output deviation and achieve instantaneous power balance; adjusting the reactive power output of the photovoltaic inverters in the target photovoltaic power station, utilizing the voltage support and reactive power regulation capabilities of the photovoltaic inverters to optimize the voltage stability and power flow distribution of the local grid without changing the active power output, indirectly alleviating the system pressure caused by output fluctuations. Conversely, if the error does not exceed the preset error value, the target photovoltaic power station is considered to be operating smoothly, and its original operating state can be maintained to avoid unnecessary disturbances. By using errors as scheduling trigger signals, a leap from reactive response to proactive intervention can be achieved. This not only significantly improves the engineering practicality and accuracy of distributed photovoltaic power output restoration, but also more effectively mitigates the impact of renewable energy fluctuations on grid voltage and frequency.

[0051] Through the above steps S102 to S110, the goal is to determine the initial power output model and optimize the initial power output model by using historical meteorological datasets and corresponding historical power generation datasets from multiple regions. The obtained target power output model is then used to determine the predicted power output of the target photovoltaic power station in the current period. Based on the actual power output of the target photovoltaic power station in the current period, the current dispatch strategy of the target photovoltaic power station is accurately determined. This achieves the technical effect of improving the accuracy of power output prediction and dispatch response efficiency of distributed photovoltaic power stations. Furthermore, it solves the technical problem in related technologies where insufficient consideration is given to the differences in meteorological conditions and meteorological observation station distribution among the various distribution areas of distributed photovoltaic power stations when dispatching distributed photovoltaic power stations in complex operating environments, resulting in low power output prediction accuracy and delayed dispatch response of distributed photovoltaic power stations.

[0052] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method. Figure 3 This is a flowchart of an optional method for determining the scheduling strategy of a distributed photovoltaic power station according to an embodiment of the present invention, as shown below. Figure 3 As shown, the method includes:

[0053] S01: Extract multiple meteorological datasets through multiple meteorological observation stations, specifically including: extracting historical meteorological datasets based on meteorological observation stations corresponding to multiple regions, and extracting historical meteorological data of each region through meteorological observation stations deployed in multiple regions to form multiple historical meteorological datasets that correspond one-to-one with multiple regions.

[0054] S02: Extract multiple power generation datasets from multiple photovoltaic power plants. Specifically, this includes: extracting historical power generation datasets corresponding to photovoltaic power plants in multiple regions, with multiple power generation datasets corresponding to multiple regions; synchronously collecting historical power generation data from distributed photovoltaic power plants corresponding to multiple regions; forming multiple historical power generation datasets that correspond one-to-one with multiple regions; ensuring that each region is associated with the power generation data of at least one photovoltaic power plant; and establishing a preliminary correspondence between meteorological conditions and power generation output.

[0055] S03: Establish multiple power generation operating conditions, specifically including: using the influence intensity of multiple historical meteorological datasets on power generation output from step S01 to obtain multiple power generation influence values. Compare each power generation influence value with a preset influence threshold. If the absolute value of the power generation influence value of a certain historical meteorological data is greater than or equal to the preset influence threshold, then the historical meteorological data is identified as a power generation influencing factor, and the current operating conditions of the power station are generated based on its historical trend; if the power generation influence values ​​corresponding to multiple historical meteorological data do not exceed the preset influence threshold, then the current operating conditions are directly determined based on the historical operating conditions of the photovoltaic power station.

[0056] S04: Conduct correlation analysis to determine the results and construct a power generation output model (initial power generation output model). Specifically, this includes: analyzing the correlation between historical meteorological datasets and multiple historical power generation datasets to determine the correlation analysis results, thereby constructing an initial power generation output model. Using the historical power generation datasets, historical meteorological datasets, and the correlation analysis results between multiple photovoltaic power plants as inputs, a distributed power generation output prediction network for multiple regions and multiple power plants is constructed. With the goal of minimizing the error between historical predicted output and actual output, iterative training is used to optimize the network parameters until the preset convergence condition is met, outputting the initialized power generation output model.

[0057] S05: Optimize the power generation output model to generate an optimized power generation output model (target power generation output model). This includes: extracting features strongly correlated with power generation output based on current operating conditions and historical power generation datasets to form a power generation output feature set; performing preprocessing on the feature set, such as standardization, missing value imputation, and irrelevant feature removal, to generate a feature transformation dataset. If the feature dimension of the feature transformation dataset exceeds a preset threshold, linear discriminant analysis is used for dimensionality reduction: samples are classified based on preset output state labels; the mean vector and covariance matrix of each class are calculated to construct inter-class scatter matrices and intra-class scatter matrices; generalized eigenvalue decomposition is performed on both, and the eigenvector corresponding to the largest eigenvalue is selected to form a projection matrix; the feature transformation dataset is projected to a low-dimensional space to generate a dimensionality-reduced dataset. If the feature dimension does not exceed the limit, the feature transformation dataset is used directly. The dimensionality-reduced dataset or the original feature transformation dataset is input into the initial power generation output model for parameter fine-tuning and structural optimization, ultimately generating a target power generation output model suitable for the current operating environment.

[0058] S06: Perform power generation output calculation and obtain power generation prediction output information (predicted power generation output), specifically including: using the target power generation output model obtained in step S05 to calculate the current power generation output of the target photovoltaic power station among multiple photovoltaic power stations, and obtain the predicted power generation output of the target photovoltaic power station in the current time period.

[0059] S07: After obtaining the actual power output information (actual power output) and predicted power output using electronic measurement equipment and calculating the error, a dispatch response strategy is formulated to improve the accuracy of photovoltaic power output restoration. Specifically, this includes: using electronic measurement equipment deployed at the target photovoltaic power station to obtain the actual power output, combining it with the predicted power output obtained in step S06 to calculate the power output error value between the two. It is then determined whether the power output error value exceeds a preset error threshold: if the error value exceeds the threshold, a compensation dispatch strategy is triggered, including but not limited to: controlling the energy storage system to perform charging and discharging operations to offset the output deviation; adjusting the active power output of other adjustable power sources in the regional power grid for power compensation; if the error value does not exceed the threshold, the current operating state of the target photovoltaic power station is maintained without intervention. The selected dispatch strategy is executed to achieve accurate restoration of distributed photovoltaic power output and grid-wide collaborative optimization.

[0060] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation method. Figure 4 This is a schematic diagram of an optional distributed photovoltaic power station scheduling strategy determination system according to an embodiment of the present invention. The system includes:

[0061] The system comprises seven modules: a meteorological dataset acquisition module 1, a power generation dataset acquisition module 2, an operating condition formulation module 3, a correlation analysis module 4, a model optimization module 5, an output calculation module 6, and an error calculation module 7. Specifically, the meteorological dataset acquisition module 1 extracts historical meteorological datasets from meteorological observation stations corresponding to multiple regions; the power generation dataset acquisition module 2 extracts historical power generation datasets from photovoltaic power plants corresponding to multiple regions, with these datasets corresponding to multiple regions; the operating condition formulation module 3 uses these historical meteorological datasets to formulate the current power generation operating conditions for multiple photovoltaic power plants; and the correlation analysis module 4 analyzes the correlation between historical meteorological datasets and these historical power generation datasets. The correlation between power data sets is determined to identify the correlation analysis results, thereby constructing an initial power generation output model. The model optimization module 5 is used to optimize the initial power generation output model based on the current power generation operating conditions and multiple historical power generation data sets to generate a target power generation output model. The output calculation module 6 is used to calculate the current power generation output of the target photovoltaic power station among multiple photovoltaic power stations using the target power generation output model to obtain the predicted power generation output of the target photovoltaic power station. The error calculation module 7 is used to obtain the actual power generation output of the target photovoltaic power station based on power measurement equipment, combine it with the predicted power generation output to perform error calculation, formulate a scheduling response strategy, and execute the scheduling response strategy to improve the accuracy of photovoltaic power generation restoration.

[0062] This embodiment also provides a scheduling strategy determination device for a distributed photovoltaic power station. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that performs a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0063] According to an embodiment of the present invention, an apparatus embodiment for implementing the above-described method for determining the scheduling strategy of a distributed photovoltaic power station is also provided. Figure 5 This is a schematic diagram of a scheduling strategy determination device for a distributed photovoltaic power station according to an embodiment of the present invention, as shown below. Figure 5 As shown, the above-mentioned distributed photovoltaic power station scheduling strategy determination device includes: a current operating condition determination module 500, an initial power generation output model construction module 502, a target power generation output model determination module 504, a predicted power generation output determination module 506, and a current scheduling strategy determination module 508, wherein:

[0064] The current operating conditions determination module 500 is used to determine the current operating conditions of each photovoltaic power station based on historical meteorological datasets of the areas where the multiple photovoltaic power stations are located.

[0065] The initial power generation model construction module 502 is connected to the current operating condition determination module 500. It is used to determine the initial power generation model based on the historical power generation datasets corresponding to multiple photovoltaic power plants and the historical meteorological datasets of the regions where the multiple photovoltaic power plants are located. The initial power generation model is used to predict the power generation output of photovoltaic power plants under different meteorological conditions.

[0066] The target power output model determination module 504 is connected to the initial power output model construction module 502. It is used to optimize the initial power output model based on the current operating conditions and historical power output datasets of multiple photovoltaic power plants to obtain the target power output model.

[0067] The predicted power generation output determination module 506 is connected to the target power generation output model determination module 504. It is used to obtain the predicted power generation output of the target photovoltaic power station in the current period based on the target power generation dataset of the target photovoltaic power station in the current period and the target meteorological dataset of the area where the target photovoltaic power station is located in the current period, using the target power generation output model.

[0068] The current scheduling strategy determination module 508 is connected to the predicted power generation output determination module 506 and is used to determine the current scheduling strategy of the target photovoltaic power station based on the actual power generation output and predicted power generation output of the target photovoltaic power station in the current time period.

[0069] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0070] It should be noted that the aforementioned current operating condition determination module 500, initial power generation output model construction module 502, target power generation output model determination module 504, predicted power generation output determination module 506, and current scheduling strategy determination module 508 correspond to steps S102 to S106 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run on a computer terminal.

[0071] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0072] The aforementioned distributed photovoltaic power station scheduling strategy determination device may also include a processor and a memory. The aforementioned current operating condition determination module 500, initial power output model construction module 502, target power output model determination module 504, predicted power output determination module 506, and current scheduling strategy determination module 508 are all stored in the memory as program modules. The processor executes the aforementioned program modules stored in the memory to realize the corresponding functions.

[0073] The processor contains a core that retrieves the corresponding program modules from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0074] According to an embodiment of this application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program runs, it controls the device where the non-volatile storage medium is located to execute any of the above-mentioned distributed photovoltaic power station scheduling strategy determination methods.

[0075] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals, and the non-volatile storage medium includes stored programs.

[0076] Optionally, during program execution, the device containing the non-volatile storage medium performs the following functions: Based on historical meteorological datasets of the regions where multiple photovoltaic power plants are located, determine the current operating conditions of each photovoltaic power plant; based on historical power generation datasets of each photovoltaic power plant and historical meteorological datasets of the regions where multiple photovoltaic power plants are located, determine an initial power generation output model, wherein the initial power generation output model is used to predict the power generation output of photovoltaic power plants under different meteorological conditions; based on the current operating conditions and historical power generation datasets of each photovoltaic power plant, optimize the initial power generation output model to obtain a target power generation output model; based on the target power generation dataset of the target photovoltaic power plant in the current time period and the target meteorological dataset of the region where the target photovoltaic power plant is located in the current time period, use the target power generation output model to obtain the predicted power generation output of the target photovoltaic power plant in the current time period; based on the actual power generation output and predicted power generation output of the target photovoltaic power plant in the current time period, determine the current scheduling strategy of the target photovoltaic power plant.

[0077] According to an embodiment of this application, an embodiment of a processor is also provided. Optionally, in this embodiment, the processor is used to run a program, wherein the program executes any of the above-described methods for determining the scheduling strategy of a distributed photovoltaic power station.

[0078] According to an embodiment of this application, an embodiment of a computer program product is also provided. Optionally, in this embodiment, the computer program product includes a computer program that, when executed by a processor, implements the steps of the method for determining the scheduling strategy of a distributed photovoltaic power station as described above.

[0079] Optionally, when the aforementioned computer program product is executed on a data processing device, it is suitable to execute an initialization program with the following method steps: determining the current operating conditions of each of the multiple photovoltaic power stations based on historical meteorological datasets of their respective regions; determining an initial power generation output model based on historical power generation datasets of each of the multiple photovoltaic power stations and historical meteorological datasets of their respective regions, wherein the initial power generation output model is used to predict the power generation output of the photovoltaic power stations under different meteorological conditions; optimizing the initial power generation output model based on the current operating conditions and historical power generation datasets of each of the multiple photovoltaic power stations to obtain a target power generation output model; obtaining the predicted power generation output of the target photovoltaic power station in the current period based on the target power generation dataset of the target photovoltaic power station in the current time period and the target meteorological dataset of the region where the target photovoltaic power station is located in the current time period, using the target power generation output model; and determining the current scheduling strategy of the target photovoltaic power station based on the actual power generation output and predicted power generation output of the target photovoltaic power station in the current period.

[0080] This invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: determining the current operating conditions of each of the multiple photovoltaic power plants based on historical meteorological datasets of their respective regions; determining an initial power generation output model based on historical power generation datasets of the multiple photovoltaic power plants and historical meteorological datasets of their respective regions, wherein the initial power generation output model is used to predict the power generation output of the photovoltaic power plants under different meteorological conditions; optimizing the initial power generation output model based on the current operating conditions and historical power generation datasets of the multiple photovoltaic power plants to obtain a target power generation output model; obtaining a predicted power generation output of the target photovoltaic power plant in the current period based on the target power generation dataset of the target photovoltaic power plant in the current time period and the target meteorological dataset of the region where the target photovoltaic power plant is located in the current time period, using the target power generation output model; and determining the current scheduling strategy of the target photovoltaic power plant based on the actual power generation output and predicted power generation output of the target photovoltaic power plant in the current period.

[0081] The order of the above embodiments of the present invention is merely for description and does not represent the superiority or inferiority of the embodiments.

[0082] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0083] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of modules described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between modules, and may be electrical or other forms.

[0084] The modules described above as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0085] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0086] If the aforementioned integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned non-volatile storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0087] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining the scheduling strategy of a distributed photovoltaic power station, characterized in that, include: Based on historical meteorological datasets of the regions where multiple photovoltaic power stations are located, the current operating conditions of each of the multiple photovoltaic power stations are determined. Based on the historical power generation datasets corresponding to each of the multiple photovoltaic power stations, and the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located, an initial power generation output model is determined, wherein the initial power generation output model is used to predict the power generation output of the photovoltaic power stations under different meteorological conditions; Based on the current operating conditions and historical power generation datasets corresponding to the multiple photovoltaic power plants, the initial power generation output model is optimized to obtain the target power generation output model; Based on the target power generation data of the target photovoltaic power station in the current time period among the multiple photovoltaic power stations, and the target meteorological data of the area where the target photovoltaic power station is located in the current time period, the target power generation output model is used to obtain the predicted power generation output of the target photovoltaic power station in the current time period. Based on the actual power output and the predicted power output of the target photovoltaic power station in the current time period, the current dispatch strategy of the target photovoltaic power station is determined.

2. The method according to claim 1, characterized in that, Before determining the current operating conditions of each of the multiple photovoltaic power stations based on historical meteorological datasets of their respective regions, the method further includes: Based on the location information of meteorological observation stations in the area where any photovoltaic power station is located, the meteorological observation range of the area where the photovoltaic power station is located is determined, wherein the location information includes at least: latitude and longitude, altitude, and the geographical environment of the area where the meteorological observation station is located; Based on a preset meteorological observation cycle, meteorological observations are conducted on the area where any photovoltaic power station is located according to the meteorological observation range to determine the initial meteorological dataset for the area where any photovoltaic power station is located. The initial meteorological dataset is optimized to obtain the historical meteorological dataset of the area where any photovoltaic power station is located; By obtaining the historical meteorological dataset of the region where any of the photovoltaic power stations is located, the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located are obtained.

3. The method according to claim 1, characterized in that, The determination of the current operating conditions of each photovoltaic power station based on historical meteorological datasets of their respective regions includes: Based on the historical meteorological dataset of the area where any photovoltaic power station is located, multiple power generation impact values ​​are determined for any photovoltaic power station. The multiple power generation impact values ​​are used to quantify the influence intensity of multiple historical meteorological data in the historical meteorological dataset of the area where any photovoltaic power station is located on the power generation output of any photovoltaic power station. The multiple power generation impact values ​​correspond one-to-one with the multiple historical meteorological data. The power generation impact value that exceeds the preset impact threshold among the multiple power generation impact values ​​is determined as the power generation impact factor of any photovoltaic power station; Based on the aforementioned factors affecting power generation, determine the current operating conditions of any of the photovoltaic power plants; or If none of the multiple power generation impact values ​​exceed the preset impact threshold, the current operating conditions of any photovoltaic power station are determined based on the historical operating conditions of any photovoltaic power station. The current operating conditions of each of the multiple photovoltaic power stations are obtained by using the method of obtaining the current operating conditions of any one of the photovoltaic power stations.

4. The method according to claim 1, characterized in that, The determination of the initial power generation model based on the historical power generation datasets corresponding to each of the multiple photovoltaic power stations, and the historical meteorological datasets of the regions where each of the multiple photovoltaic power stations is located, includes: Based on the historical power generation datasets corresponding to each of the multiple photovoltaic power stations, and the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located, multiple historical correlation results are determined. The multiple historical correlation results correspond one-to-one with the multiple photovoltaic power stations, and the multiple historical correlation results are used to quantify the degree of correlation between the historical power generation datasets of the corresponding photovoltaic power stations and the historical meteorological datasets of the corresponding regions. Based on the historical power generation datasets corresponding to each of the multiple photovoltaic power stations, the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located, the multiple historical correlation results, and the actual power generation output of each of the multiple photovoltaic power stations during the historical prediction period, the power generation output network is optimized and trained with the goal of minimizing the difference between any actual power generation output and the corresponding predicted power generation output, until a predetermined number of iterations is reached. The corresponding predicted historical power generation output is obtained by using the power generation output network to predict based on the historical power generation datasets, historical meteorological datasets, and historical correlation results corresponding to any of the multiple photovoltaic power stations during the optimization training process. The initial power generation model is obtained based on the optimized power generation network obtained when the predetermined number of iterations is reached.

5. The method according to claim 1, characterized in that, The optimization of the initial power output model based on the current operating conditions and historical power generation datasets of the multiple photovoltaic power plants to obtain the target power output model includes: Based on the current operating conditions and historical power generation datasets corresponding to the multiple photovoltaic power plants, a power generation output feature set is obtained, wherein the power generation output feature set is a set of features associated with the power generation output of the multiple photovoltaic power plants; The power generation output feature set is subjected to feature transformation to obtain a feature transformation dataset, wherein the feature transformation is used to convert the power generation output feature set into a dataset that conforms to the input dimension of the initial power generation output model; If the number of features in the feature transformation dataset exceeds a preset number, the feature transformation dataset is subjected to dimensionality reduction processing to obtain a dimensionality-reduced dataset; based on the dimensionality-reduced dataset, the initial power generation output model is optimized to obtain the target power generation output model; or If the number of features in the feature transformation dataset does not exceed the preset number, the initial power generation output model is optimized based on the feature transformation dataset to obtain the target power generation output model.

6. The method according to claim 5, characterized in that, When the number of features in the feature transformation dataset exceeds a preset number, the feature transformation dataset is subjected to dimensionality reduction processing to obtain a dimensionality-reduced dataset, including: When the number of features in the feature transformation dataset exceeds the preset number, multiple feature data in the feature transformation dataset are classified based on multiple preset category labels to obtain multiple classification subsets. The multiple preset category labels include at least a high-output state label, a medium-output state label, and a low-output state label. The high-output state label is used to identify feature data whose power generation output is greater than or equal to a preset first threshold. The medium-output state label is used to identify feature data whose power generation output is between the preset first threshold and a preset second threshold. The low-output state label is used to identify feature data whose power generation output is less than or equal to the preset second threshold. The preset first threshold is greater than the preset second threshold. Each of the multiple classification subsets includes feature data corresponding to a preset category label. Determine the mean vector and covariance matrix corresponding to each of the multiple classification subsets, wherein any element in the mean vector is the average value of the samples included in the corresponding feature data in the corresponding classification subset; any off-diagonal element in the covariance matrix is ​​the covariance between the samples included in the two feature data in the corresponding classification subset, and any diagonal element is the variance of the samples included in the corresponding feature data in the corresponding classification subset. Based on the mean vectors corresponding to each of the multiple classification subsets, an inter-class scatter matrix is ​​obtained, wherein the inter-class scatter matrix is ​​used to quantify the degree of dispersion among the multiple classification subsets. Based on the covariance matrix corresponding to each of the multiple classification subsets and the total number of samples corresponding to each of the multiple classification subsets, an intra-class scatter matrix is ​​obtained, wherein the intra-class scatter matrix is ​​used to quantify the degree of dispersion among multiple feature data in any one of the multiple classification subsets. The inter-class scatter matrix and the intra-class scatter matrix are subjected to generalized eigenvalue decomposition to obtain a target generalized eigenvector, wherein the target generalized eigenvector represents the generalized eigenvector corresponding to the generalized eigenvalue greater than a preset eigenvalue threshold among a plurality of generalized eigenvalues; When there are multiple target generalized feature vectors, the multiple target generalized feature vectors are juxtaposed in the column direction based on a preset sorting method to obtain a feature vector projection matrix. The preset sorting method is the sorting of the generalized feature values ​​corresponding to the multiple target generalized feature vectors from largest to smallest. Based on the feature vector projection matrix, the feature transformation dataset is subjected to dimensionality reduction processing to obtain the dimensionality-reduced dataset.

7. The method according to any one of claims 1 to 6, characterized in that, The step of determining the current dispatch strategy for the target photovoltaic power station based on its actual power generation and predicted power generation during the current time period includes: Based on the actual power generation output and the predicted power generation output, the power generation output error value of the target photovoltaic power station is determined; If the power generation error value exceeds a preset error value, the current scheduling strategy is determined to be compensatory scheduling of the target photovoltaic power station. The compensatory scheduling includes at least one of the following: controlling energy storage devices to perform charging and discharging operations to compensate for the power generation error value; or adjusting the reactive power output of the target photovoltaic power station to compensate for the power generation error value; or If the power generation error value does not exceed the preset error value, the current scheduling strategy is determined to maintain the current operating state of the target photovoltaic power station.

8. A device for determining the scheduling strategy of a distributed photovoltaic power station, characterized in that, include: The current operating conditions determination module is used to determine the current operating conditions of each of the multiple photovoltaic power stations based on historical meteorological datasets of the respective regions of the multiple photovoltaic power stations. The initial power generation model construction module is used to determine the initial power generation model based on the historical power generation datasets corresponding to each of the multiple photovoltaic power stations and the historical meteorological datasets of the regions where the multiple photovoltaic power stations are located. The initial power generation model is used to predict the power generation output of the photovoltaic power stations under different meteorological conditions. The target power output model determination module is used to optimize the initial power output model based on the current operating conditions and historical power output datasets corresponding to the multiple photovoltaic power plants to obtain the target power output model. The predictive power generation output determination module is used to obtain the predicted power generation output of the target photovoltaic power station in the current period based on the target power generation dataset of the target photovoltaic power station in the current period and the target meteorological dataset of the area where the target photovoltaic power station is located in the current period, using the target power generation output model. The current scheduling strategy determination module is used to determine the current scheduling strategy of the target photovoltaic power station based on the actual power generation output and the predicted power generation output of the target photovoltaic power station in the current time period.

9. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores multiple instructions, which are adapted to be loaded and executed by a processor to determine the scheduling strategy of a distributed photovoltaic power station according to any one of claims 1 to 7.

10. An electronic device, characterized in that, It includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method for determining the scheduling strategy of a distributed photovoltaic power station as described in any one of claims 1 to 7.

Images