A photovoltaic output fluctuation risk quantification method and system based on an unsupervised network

Abstract

The application relates to a photovoltaic output fluctuation risk quantification method and system based on an unsupervised network, and belongs to the technical field of power system control. The method comprises the following steps: acquiring historical output data of a photovoltaic power station and constructing a multi-dimensional feature vector representing intra-day fluctuation characteristics; based on the feature vector, an improved dynamic time warping distance is used to perform unsupervised clustering on daily output curves to obtain typical output modes; fluctuation risk events are identified based on an adaptive threshold value, and event attributes are extracted; a fluctuation event complex network is constructed, the edge weight of the network is calculated in multiple dimensions by coupling time intervals, direction cooperativity and intensity coupling factors, and the spatiotemporal clustering characteristics of the risk are quantified based on network topology indexes; the fluctuation risk event density in each typical mode is calculated, and the quantitative correlation between the mode and the risk is established; and finally, multi-time-scale statistical analysis and short-time-sequence mode mining are performed.

Description

Technical Field

[0001] This application relates to the field of operation analysis and control technology of new energy power systems, and more specifically, to a method and system for quantifying photovoltaic output fluctuation risk based on unsupervised networks. Background Technology

[0002] With the increasing penetration rate of photovoltaic (PV) power generation, the inherent intermittency and volatility of its output pose a severe challenge to the safe and stable operation of the power system. PV output is affected by various meteorological factors such as sunlight, temperature, and cloud cover, exhibiting strong randomness and uncertainty, which may lead to problems such as grid frequency deviation, line over-limit, and even voltage instability. Therefore, accurately quantifying the volatility risk of PV output is of great significance for the planning, operation, and proactive defense of the power grid.

[0003] In existing technologies, photovoltaic (PV) power output fluctuation risk assessment methods are mainly divided into two categories. The first category is meteorological data-driven methods, which predict fluctuations by establishing physical or statistical models between meteorological factors and PV power output. However, these methods heavily rely on high-precision, high-coverage meteorological monitoring data, resulting in high data acquisition costs and limited applicability in complex terrain or micro-meteorological environments. Furthermore, the spatiotemporal resolution of meteorological data often does not match that of PV power output data, potentially introducing errors. The second category is statistical methods based on pure power output data, which mainly assess risk by calculating statistical indicators such as historical power output fluctuation rates and ramp-up rates. While these methods avoid dependence on external data, they are mostly single-point or simple statistical analyses, failing to deeply explore the inherent correlations and clustering characteristics of fluctuation events in the temporal and spatial dimensions. For example, they cannot effectively identify whether fluctuation events tend to occur in clusters or whether there are key risk propagation paths or sources. Existing methods struggle to self-organize typical power output patterns from massive amounts of historical data, quantify the spatiotemporal cluster structure of fluctuation risks, and reveal their deep correlation with operating modes. This results in relatively one-sided risk assessment results, a lack of insight into the risk formation mechanism, and an inability to provide sufficient basis for the refined scheduling of the power grid.

[0004] Therefore, there is an urgent need for a technical solution that does not rely on external meteorological data, can automatically extract the inherent laws of fluctuations from the historical output data of photovoltaic power generation, and can systematically quantify its spatiotemporal cluster risk characteristics, so as to make up for the shortcomings of existing technologies. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a method and system for quantifying photovoltaic power output fluctuation risk based on unsupervised networks.

[0006] The technical solution of this invention is as follows: This invention proposes a method for quantifying the risk of photovoltaic power output fluctuations based on unsupervised networks, comprising the following steps: Acquire historical power output time-series data of photovoltaic power plants and construct a multi-dimensional daily feature vector to characterize intraday power output fluctuations. Based on multidimensional daily feature vectors, an improved dynamic time warping distance is used as a similarity measure to cluster intraday power output curves and obtain typical daily power output patterns. Based on historical output time series data, the fluctuation risk events are identified through adaptive thresholds, and the start time, duration, fluctuation amplitude and event type of each fluctuation risk event are extracted. Based on the extracted volatility risk events, a complex network of volatility events is constructed. Based on the complex network of volatility events, network topology indicators are calculated to quantify the spatiotemporal clustering characteristics of volatility risk. Based on typical daily power output patterns and extracted volatility risk events, the volatility risk event density under each typical power output pattern is calculated through a complex network of volatility events, and a quantitative correlation between typical power output patterns and volatility risk levels is established. Historical output time series data are aggregated to different time scales, and statistical characteristics at each scale are calculated to analyze the scaling effect. At the same time, based on historical output time series data, continuous hourly output sequences associated with volatility risk events are extracted and clustered to discover frequently occurring output time series patterns and calculate the transition probability between different patterns.

[0007] Preferably, the multidimensional daily feature vector includes: sunrise power pattern dispersion and sunrise power fluctuation path complexity; the sunrise power pattern dispersion is obtained by calculating the Shannon entropy of the intraday power output distribution; the sunrise power fluctuation path complexity is obtained by calculating the approximate entropy of the intraday power output sequence.

[0008] Preferably, the improved dynamic time warping distance adds a fluctuation trend penalty factor to the alignment points with opposite power change directions when calculating the warping path of the two output curves.

[0009] Preferably, in the complex network of volatility events, the nodes represent volatility risk events, and the event types of the volatility risk events include rising volatility events and falling volatility events.

[0010] Preferably, in the complex network of volatility events, an edge is established when the time interval between two volatility risk events is less than a set threshold, and the formula for calculating the edge weight is: ; In the formula: W is the edge weight; The exponential decay factor is the time interval between two volatility risk events; The coefficient representing the directional coordination of fluctuations; This is the wave intensity coupling factor.

[0011] Preferably, when constructing a complex network of regional fluctuation events involving multiple photovoltaic power plants, the influencing factors of the edge weights also include the geographical cluster potential factor, the value of which is negatively correlated with the electrical distance between the power plants involved in the two fluctuation risk events.

[0012] Preferably, the network topology metrics include the average clustering coefficient, node weighted degree centrality, and network modularity.

[0013] On the other hand, the present invention also provides a photovoltaic power output fluctuation risk quantification system based on unsupervised networks, comprising: The data acquisition module obtains historical power output time-series data of photovoltaic power plants and constructs a multi-dimensional daily feature vector to characterize the intraday power output fluctuation characteristics. The unsupervised clustering module, based on multi-dimensional daily feature vectors, uses an improved dynamic time warping distance as a similarity measure to cluster intraday power output curves, thereby obtaining typical daily power output patterns. The volatility risk event identification module identifies volatility risk events based on historical output time series data through adaptive thresholds, and extracts the start time, duration, volatility amplitude, and event type of each volatility risk event. The complex network construction module constructs a complex network of volatility events based on the extracted volatility risk events, and calculates network topology indicators based on the complex network of volatility events to quantify the spatiotemporal cluster characteristics of volatility risk. The module analyzing the correlation mechanism between volatility risk and output mode, based on typical daily output modes and extracted volatility risk events, calculates the volatility risk event density under each typical output mode through a complex network of volatility events, and establishes a quantitative correlation between typical output modes and volatility risk levels. The multi-timescale statistics and sequence pattern mining module aggregates historical power output time series data to different time scales, calculates statistical characteristics at each scale to analyze the scale effect; at the same time, based on historical power output time series data, it extracts continuous hourly power output sequences associated with volatility risk events for clustering, mines frequently occurring power output time series patterns, and calculates the transition probability between different patterns.

[0014] In another aspect, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a photovoltaic power output fluctuation risk quantification method based on an unsupervised network as described in any embodiment of the present invention.

[0015] In another aspect, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for quantifying photovoltaic power output fluctuation risk based on an unsupervised network as described in any embodiment of the present invention.

[0016] The present invention has the following beneficial effects: 1. This invention is based entirely on the historical output data of the photovoltaic power station itself for mining and analysis. It does not require the introduction of external meteorological parameters such as irradiance and temperature, which significantly reduces the cost and complexity of data acquisition and improves the universality and engineering applicability of the method. It is especially suitable for scenarios where meteorological monitoring data is insufficient or of low quality.

[0017] 2. By using dynamic time warping distance with a fluctuation trend penalty factor as a similarity measure for clustering, the similarity judgment of the daily power output curves not only considers the shape and outline, but also pays more attention to the synergy of fluctuation trends. This allows for a more accurate division of typical power output patterns that reflect different operating states and risk levels, laying a more reliable foundation for subsequent risk correlation analysis.

[0018] 3. By constructing a complex network of volatile events and designing a multi-dimensional edge weight calculation formula that integrates time decay, directional coordination, and strength coupling, the abstract volatility risk is transformed into a visible and measurable network topology. By calculating indicators such as the network's average clustering coefficient and node weighted degree centrality, it is possible to effectively reveal whether volatile events erupt in a concentrated manner in time and whether they propagate in a correlated manner in space. It also identifies core risk events and high-risk event clusters in the network, achieving a fine characterization of the "clustering" and "propagation" of risks, far exceeding the analytical dimensions of traditional statistical methods.

[0019] 4. This invention not only analyzes daily-level typical patterns and event-level networks, but also conducts statistical characteristic analysis at multiple time scales and short-term behavioral pattern mining at the hourly level. It can comprehensively evaluate the variation law of fluctuation characteristics with the observation scale and capture the evolution law of power output form in a short period of time before and after fluctuation events. This provides richer input information and physical constraints for ultra-short-term power prediction and frequency control, and improves the systematic nature of the method. Attached Figure Description

[0020] Figure 1 This is a flowchart of the method of the present invention; Figure 2 Clustering results and seasonal distribution map of typical daily power generation patterns; Figure 3 A time distribution characteristic diagram of volatility risk events; Figure 4 A schematic diagram of the topology of a complex network of fluctuating events; Figure 5 The curves show the variation of the output coefficient of variation at different time scales. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0022] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0023] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0024] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0025] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0026] Example 1: To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present application and with reference to the accompanying drawings.

[0027] To address the problems of existing technologies, this invention provides a method for quantifying the risk of photovoltaic power output fluctuations based on unsupervised networks, comprising the following steps: Acquire historical power output time-series data of photovoltaic power plants and construct a multi-dimensional daily feature vector to characterize intraday power output fluctuations. To obtain historical power output time-series data of the target photovoltaic power station, this embodiment uses the actual power output data (i.e., time-series data) of a photovoltaic power station in a certain year with a sampling interval of 5 minutes as an example. The rated power of the photovoltaic power station... The capacity is 100MW. First, the output data is preprocessed, including: removing abnormal zero or negative values ​​caused by sensor malfunctions; for missing data points, linear interpolation of adjacent time points is used to fill in the missing data points.

[0028] Based on cleaned and corrected daily data, a multidimensional daily feature vector is constructed to characterize intraday power output fluctuations. This feature vector includes, but is not limited to, the following two core features: Daily power pattern dispersion: The normalized daily power curve is divided into 10 equal intervals based on power value. The probability of the power value distribution for that day is calculated, and then its Shannon entropy is calculated. This entropy value quantifies the degree of dispersion of daily power across the entire power range. The higher the entropy value, the more uncertain the daily power pattern is, and the greater the potential fluctuation risk.

[0029] Daily power output fluctuation path complexity: The approximate entropy is calculated using the daily normalized power output time series (288 points in total) as input. During calculation, the embedding dimension m is set to 2, and the similarity tolerance r is set to 0.2 times the standard deviation of the series. This approximate entropy value quantifies the unpredictability and pattern complexity of short-term fluctuations in the power output curve.

[0030] The basic characteristics include: daily maximum power, daily average power, and further derived "intraday volatility" (defined as the standard deviation of the power output difference between adjacent time points within a day) and "maximum ramp rate".

[0031] Based on multidimensional daily feature vectors, an improved dynamic time warping distance is used as a similarity measure to cluster intraday power output curves and obtain typical daily power output patterns. The daily output curves throughout the year are normalized to eliminate the influence of absolute power magnitude and focus on curve shape. An improved Dynamic Time Warping (DTW) distance is used as a metric to measure the similarity between two output curves.

[0032] The improvement lies in: when calculating the DTW regularized path of two curves A and B, for each pair of alignment points on the path ( , ), not only calculating its Euclidean distance as the base distance. It also determines the power change trend of that point in its local area. If the point... The point is located on the rising segment of its curve. If the curve is in the descending segment (or vice versa), then the two are considered to have opposite fluctuation trends, and a penalty weight is added to the basic distance. (This embodiment takes) =1.2). That is, the final distance between the two points. This improvement allows clustering to focus more on the synergy of fluctuation trends.

[0033] Based on the improved DTW distance, the K-medoids clustering algorithm was used to cluster the sunrise power curves throughout the year. By calculating the silhouette coefficients of the clustering results under different K values ​​(K=2 to 10), the optimal number of clusters, K=7, was determined. Specifically, as follows... Figure 2 As shown, the annual daily power output curve is finally divided into 7 typical patterns. For example, pattern 1 is a stable power output pattern that dominates throughout the year, while patterns 6 and 7 are high-output, high-fluctuation patterns unique to summer.

[0034] Based on historical output time series data, the fluctuation risk events are identified through adaptive thresholds, and the start time, duration, fluctuation amplitude and event type of each fluctuation risk event are extracted. In this embodiment, the adaptive threshold Using this threshold as a benchmark, scan the 5-minute resolution power output data, identify all consecutive power increases or decreases exceeding this threshold, and define them as a volatility risk event.

[0035] like Figure 3 As shown, for each identified event, four key attributes are extracted: start time, duration (minutes), fluctuation range (MW), and event type. The event type is defined as follows: events with continuously rising power are "rising fluctuation events," and events with continuously falling power are "falling fluctuation events." In the annual data of this embodiment, a total of 1168 fluctuation risk events were identified.

[0036] Based on the extracted volatility risk events, a complex network of volatility events is constructed. Based on the complex network of volatility events, network topology indicators are calculated to quantify the spatiotemporal clustering characteristics of volatility risk. In this embodiment, as Figure 4 As shown, a complex network of volatility events is constructed based on all volatility risk events. The network construction rules are as follows: Node: Each volatility risk event is considered a node.

[0037] Edge connection condition: If the time interval between two volatility risk events is less than a preset threshold T (T=60 minutes in this embodiment), then an edge is established between the two corresponding nodes.

[0038] Edge weight calculation: The edge weight W is a multi-dimensional coupling value, and the calculation formula is as follows: ; In the formula: W is the edge weight; This is the exponential decay factor for the time interval between two volatility risk events. , The time interval between two volatility risk events. The time decay constant is set to 1 in this embodiment; The coefficient represents the directional coordination of fluctuations. If two events are of the same type, the first coefficient value is used; if they are opposite, the second coefficient value is used. The first coefficient value is greater than the second coefficient value. In this embodiment, the first coefficient value is 1.5 and the second coefficient value is 0.5. This is the volatility intensity coupling factor, whose value is positively correlated with the product of the volatility amplitudes of the two events, specifically: , These represent the volatility amplitudes of the two volatility risk events. This represents the largest fluctuation range observed in historical data. This factor emphasizes that the correlation between large fluctuation events is more risky.

[0039] As a preferred embodiment of this practice, for multi-power station area analysis, a geographical clustering potential factor can be introduced into the above formula. ,Right now . The value of is negatively correlated with the electrical distance between the power stations to which the two events belong; the closer the electrical distance, the better. The larger the value, the more specific it is. d is the electrical distance, and D is the distance attenuation constant.

[0040] Based on the constructed weighted undirected network, network topology indicators are calculated to quantify the spatiotemporal clustering characteristics of volatility risk. The indicators calculated in this embodiment include: Average clustering coefficient of the network: measures the tendency of events to form clusters locally. The calculated average clustering coefficient of this network is 0.512, indicating that volatility risk events have significant spatiotemporal clustering.

[0041] Node weighted degree centrality: Identifying core risk events (hub nodes) in the network.

[0042] Network modularity: used to discover community structures in a network and identify clusters of high-risk events.

[0043] Based on typical daily power output patterns and extracted volatility risk events, the volatility risk event density under each typical power output pattern is calculated through a complex network of volatility events, and a quantitative correlation between typical power output patterns and volatility risk levels is established. The total number of volatility risk events occurring within all days belonging to each typical daily power pattern is counted. This total number of events is then divided by the total number of days in the corresponding pattern to obtain the "volatility risk event density" for each typical pattern.

[0044] Analysis revealed that Mode 7 (high-output summer mode) had the highest risk event density at 3.570 events / day, followed by Mode 6 at 3.274 events / day. Mode 1 (stable mode) had the lowest risk density. This quantitatively reveals the inherent mechanism by which "high-output operating modes are often accompanied by higher volatility risk."

[0045] Historical output time series data are aggregated to different time scales, and statistical characteristics at each scale are calculated to analyze the scaling effect. At the same time, based on historical output time series data, continuous hourly output sequences associated with volatility risk events are extracted and clustered to discover frequently occurring output time series patterns and calculate the transition probability between different patterns.

[0046] The original 5-minute resolution power output data was aggregated to 15-minute, 30-minute, and 60-minute time scales, and the mean, standard deviation, and coefficient of variation of the power output series were calculated for each scale. Analysis revealed that as the time scale increased from 5 minutes to 60 minutes, the coefficient of variation of the power output series slowly decreased from 1.727 to 1.701, indicating that photovoltaic power output exhibits strong volatility across different time scales, and while a scale effect exists, the changes are not drastic within this range. Specifically... Figure 5 As shown.

[0047] Simultaneously, continuous hourly output sequences associated with identified volatility risk events are extracted from historical data (e.g., taking several hours before and after the event's occurrence). After normalizing these sequences, cluster analysis (e.g., using the K-means algorithm) is performed again to uncover frequently occurring short-term output patterns (e.g., the "rapid rise-plateau-gradual fall" pattern). Furthermore, by calculating the state transition probability matrix, the evolutionary patterns among these short-term patterns are analyzed; for example, it is found that a certain baseline pattern appears most frequently, and other patterns tend to "regress" to this baseline pattern.

[0048] Example 2: This embodiment provides a photovoltaic power output fluctuation risk quantification system based on unsupervised networks, including: The data acquisition module obtains historical power output time-series data of photovoltaic power plants and constructs a multi-dimensional daily feature vector to characterize the intraday power output fluctuation characteristics. The unsupervised clustering module, based on multi-dimensional daily feature vectors, uses an improved dynamic time warping distance as a similarity measure to cluster intraday power output curves, thereby obtaining typical daily power output patterns. The volatility risk event identification module identifies volatility risk events based on historical output time series data through adaptive thresholds, and extracts the start time, duration, volatility amplitude, and event type of each volatility risk event. The complex network construction module constructs a complex network of volatility events based on the extracted volatility risk events, and calculates network topology indicators based on the complex network of volatility events to quantify the spatiotemporal cluster characteristics of volatility risk. The module analyzing the correlation mechanism between volatility risk and output mode, based on typical daily output modes and extracted volatility risk events, calculates the volatility risk event density under each typical output mode through a complex network of volatility events, and establishes a quantitative correlation between typical output modes and volatility risk levels. The multi-timescale statistics and sequence pattern mining module aggregates historical power output time series data to different time scales, calculates statistical characteristics at each scale to analyze the scale effect; at the same time, based on historical power output time series data, it extracts continuous hourly power output sequences associated with volatility risk events for clustering, mines frequently occurring power output time series patterns, and calculates the transition probability between different patterns.

[0049] Example 3: This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a photovoltaic power output fluctuation risk quantification method based on an unsupervised network as described in any embodiment of the present invention.

[0050] Example 4: This embodiment proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements a photovoltaic power output fluctuation risk quantification method based on an unsupervised network as described in any embodiment of the present invention.

[0051] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0052] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0053] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0054] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a 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 described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks, characterized in that, Includes the following steps: Acquire historical power output time-series data of photovoltaic power plants and construct a multi-dimensional daily feature vector to characterize intraday power output fluctuations. Based on multidimensional daily feature vectors, an improved dynamic time warping distance is used as a similarity measure to cluster intraday power output curves and obtain typical daily power output patterns. Based on historical output time series data, the fluctuation risk events are identified through adaptive thresholds, and the start time, duration, fluctuation amplitude and event type of each fluctuation risk event are extracted. Based on the extracted volatility risk events, a complex network of volatility events is constructed. Based on the complex network of volatility events, network topology indicators are calculated to quantify the spatiotemporal clustering characteristics of volatility risk. Based on typical daily power output patterns and extracted volatility risk events, the volatility risk event density under each typical power output pattern is calculated through a complex network of volatility events, and a quantitative correlation between typical power output patterns and volatility risk levels is established. Historical output time series data are aggregated to different time scales, and statistical characteristics at each scale are calculated to analyze the scaling effect. Meanwhile, based on historical power output time series data, continuous hourly power output sequences associated with volatility risk events are extracted and clustered to uncover frequently occurring power output time series patterns and calculate the transition probability between different patterns.

2. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 1, characterized in that: The multidimensional daily feature vector includes: the dispersion of the sunrise power pattern and the complexity of the sunrise power fluctuation path; the dispersion of the sunrise power pattern is obtained by calculating the Shannon entropy of the intraday power distribution; the complexity of the sunrise power fluctuation path is obtained by calculating the approximate entropy of the intraday power sequence.

3. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 1, characterized in that: The improved dynamic time warping distance adds a fluctuation trend penalty factor to the alignment points with opposite power change directions when calculating the warping path of the two output curves.

4. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 1, characterized in that: In the complex network of volatility events, nodes represent volatility risk events, and the types of volatility risk events include rising volatility events and falling volatility events.

5. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 1, characterized in that: In the complex network of volatility events, an edge is established when the time interval between two volatility risk events is less than a set threshold, and the formula for calculating the edge weight is as follows: ； In the formula: W is the edge weight; The exponential decay factor is the time interval between two volatility risk events; The coefficient representing the directional coordination of fluctuations; This is the wave intensity coupling factor.

6. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 5, characterized in that: When constructing a complex network of regional fluctuation events involving multiple photovoltaic power plants, the factors influencing the edge weights also include the geographical cluster potential factor, the value of which is negatively correlated with the electrical distance between the power plants involved in the two fluctuation risk events.

7. The method for quantifying photovoltaic power output fluctuation risk based on unsupervised networks according to claim 1, characterized in that: The network topology metrics include the average clustering coefficient, node weighted degree centrality, and network modularity.

8. A photovoltaic power output fluctuation risk quantification system based on unsupervised networks, characterized in that, include: The data acquisition module obtains historical power output time-series data of photovoltaic power plants and constructs a multi-dimensional daily feature vector to characterize the intraday power output fluctuation characteristics. The unsupervised clustering module, based on multi-dimensional daily feature vectors, uses an improved dynamic time warping distance as a similarity measure to cluster intraday power output curves, thereby obtaining typical daily power output patterns. The volatility risk event identification module identifies volatility risk events based on historical output time series data through adaptive thresholds, and extracts the start time, duration, volatility amplitude, and event type of each volatility risk event. The complex network construction module constructs a complex network of volatility events based on the extracted volatility risk events, and calculates network topology indicators based on the complex network of volatility events to quantify the spatiotemporal cluster characteristics of volatility risk. The module analyzing the correlation mechanism between volatility risk and output mode, based on typical daily output modes and extracted volatility risk events, calculates the volatility risk event density under each typical output mode through a complex network of volatility events, and establishes a quantitative correlation between typical output modes and volatility risk levels. The multi-timescale statistics and sequence pattern mining module aggregates historical power output time series data to different time scales, calculates statistical characteristics at each scale to analyze the scale effect; at the same time, based on historical power output time series data, it extracts continuous hourly power output sequences associated with volatility risk events for clustering, mines frequently occurring power output time series patterns, and calculates the transition probability between different patterns.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a photovoltaic power output fluctuation risk quantification method based on unsupervised networks as described in claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements a photovoltaic power output fluctuation risk quantification method based on unsupervised networks as described in claims 1-7.

Images