Photovoltaic irradiance prediction method based on trend deduction

Abstract

The invention belongs to the technical field of power distribution automation, and particularly relates to a photovoltaic irradiance prediction method based on trend deduction. Comprising the following steps: S1, data acquisition and gridding deployment; s2, constructing an ideal irradiance curve; s3, identifying and processing rain and snow weather influence; s4, deducing the change trend of the cloud layer; and S5, predicting the regional irradiance. According to the method, the discretely distributed micro-meteorological sensor data and the historical irradiance rule are fused, a photovoltaic irradiance prediction model considering space-time evolution is constructed, and high-precision prediction from point observation to surface deduction is realized.

Description

Technical Field

[0001] This invention belongs to the field of power distribution automation technology, specifically relating to a photovoltaic irradiance prediction method based on trend extrapolation. Background Technology

[0002] Driven by green and low-carbon goals, photovoltaic (PV) power generation capacity has experienced explosive growth, providing a large amount of clean energy to the power system. However, PV power output is significantly intermittent and fluctuating, and its power generation capacity is highly dependent on surface solar irradiance, posing a severe challenge to the stable operation and dispatch of the power grid. Therefore, high-precision prediction of PV irradiance has become one of the key technologies for improving PV absorption capacity and ensuring the safe operation of the power grid.

[0003] The most fundamental aspect of predicting photovoltaic power generation capacity is the prediction of irradiance at different latitudes and longitudes, which traditional prediction methods struggle to accurately assess. Firstly, cloud formations affecting irradiance are three-dimensional, and atmospheric conditions at different altitudes influence irradiance. Secondly, solar irradiance exhibits different characteristics at different latitudes and longitudes. Finally, irradiance is influenced by multiple factors, including season, time, wind force, and direction, resulting in complex variation patterns. Summary of the Invention

[0004] The purpose of this invention is to provide a photovoltaic irradiance prediction method based on trend extrapolation, so as to solve the problems existing in the background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic irradiance prediction method based on trend extrapolation, comprising the following steps:

[0006] S1. Data Acquisition and Gridded Deployment: Within the target prediction area, multiple observation points are set at preset intervals, and micro-meteorological sensors are deployed at each observation point to collect micro-meteorological data, including irradiance, wind force, and wind direction, at a first preset frequency.

[0007] S2. Constructing the ideal irradiance curve: Based on the historical irradiance data of the target prediction area, extract the ideal irradiance curves for different seasons under sunny weather conditions;

[0008] S3. Identify and process the impact of rain and snow weather: Based on meteorological forecast data, identify the types of rain and snow weather, and set the irradiance forecast value for the corresponding weather type within the forecast period to zero;

[0009] S4. Cloud Layer Change Trend Prediction: For non-rainy / snowy weather, based on the aforementioned micrometeorological data, combined with the cloud layer blocking rate model and spatiotemporal extrapolation algorithm, the cloud layer blocking rate change trend at various locations in the future period is predicted; the cloud layer blocking rate B is calculated using the formula... Calculate, where R represents the current irradiance and Rm represents the ideal irradiance at this point in time;

[0010] S5. Regional Irradiance Prediction: By integrating the ideal irradiance curve with the predicted cloud blocking rate, the predicted irradiance values ​​for each location within the target prediction area in the future time period are calculated.

[0011] Preferably, constructing the ideal irradiance curve in step S2 specifically includes:

[0012] S21. Group the historical irradiance data according to geographical coordinates and season;

[0013] S22. For each group, using the irradiance curve of the day with the longest irradiance duration as the benchmark, the irradiance curves of other days are time-stretched and aligned with the midday point as the center of symmetry.

[0014] S23. Filter out data points with sunny weather from the aligned curves and remove outliers;

[0015] S24. Interpolate the processed data to generate the ideal irradiance curve in minutes.

[0016] Preferably, in step S4, the cloud change trend inference includes cloud transfer inference based on near-surface wind force and direction and autoregressive inference based on local historical irradiance data.

[0017] The cloud transfer extrapolation based on near-surface wind force and direction: Calculate the cloud origin trajectory sequence {P1,P2,...,P} based on wind force and direction. n}, and through formula Predict the blocking rate B1 at location P in the next n minutes, where λ1 is the cloud attenuation coefficient per kilometer, and B(P i ) represents position P i The current blocking rate;

[0018] Autoregressive extrapolation based on local historical irradiance data: After logarithmically transforming the historical blocking rate data, the autoregressive model is used to predict the blocking rate B2 in the next n minutes.

[0019] Preferably, in step S4, the final position P after n minutes is calculated using the formula B = B1 × λ2 + B2 × (1 - λ2), where λ2 is a weighting coefficient, the value of which is determined by optimizing the residuals verified by historical data.

[0020] Preferably, before step S5, step S0, a full-map simulation of the current meteorological conditions, is also included:

[0021] S01. Divide the target prediction area into a dense grid with a grid length of 1 kilometer;

[0022] S02. Based on the real-time obstruction rate of each observation point, the k-nearest neighbor method is used to find the three nearest observation points for each grid point, and the formula is used to... Calculate the initial blocking rate for each grid point, where L(P,P) i ) represents P, P i The distance between two points;

[0023] S03. Starting from the initial blocking rate distribution map, execute step S4 to deduce the blocking rate distribution at future times.

[0024] Preferably, in step S03, the deduced blocking rate B4 at time t and the initial blocking rate B3 are fused using the formula B=B3×λ3+B4×(1-λ3), where λ3 is a weighting coefficient, the value of which is determined by verification residual optimization.

[0025] Preferably, in step S5, the formula E(P,t)=B(P,t)×E is used. m The irradiance prediction E(P,t) for location P at future time t is calculated, where B(P,t) represents the predicted blocking rate of location P at future time t, and E... m (P,t) represents the calculated ideal irradiance value for location P at a future time t.

[0026] Preferably, the first preset frequency is once every 5 minutes, the future time period is 2 hours, the preset spacing is 10 kilometers, and the observation points are deployed within a radius of 0.5 kilometers centered on the grid points.

[0027] The beneficial effects of this invention are as follows: By integrating discretely distributed micro-meteorological sensor data with historical irradiance patterns, a photovoltaic irradiance prediction model considering spatiotemporal evolution is constructed, achieving high-precision prediction from "point" observation to "area" extrapolation; by introducing the concept of cloud blocking rate and integrating a cloud transfer model driven by near-surface wind force and direction with local change trend analysis based on autoregression, the dynamic characteristics of cloud movement are effectively captured, significantly improving the accuracy and reliability of short-term (future 2 hours) irradiance prediction under complex weather conditions such as cloudy, rain, and snow; ultimately, it provides accurate data support for photovoltaic power generation prediction, enhances the grid's ability to absorb photovoltaic fluctuations, and effectively ensures the safe and stable operation of the power system. Attached Figure Description

[0028] Figure 1 This is the logic diagram for predicting irradiance in this invention. Detailed Implementation

[0029] The present invention will be further described in detail below with reference to specific embodiments. The core of the present invention lies in achieving high-precision prediction of regional photovoltaic irradiance within the next 2 hours by combining discretely distributed micro-meteorological sensor data with historical patterns and real-time trends.

[0030] S1. Deployment and Data Acquisition of Micro-meteorological Sensor Network

[0031] Determining the location of observation points: Taking the target area as the objective and using a step size of 10 kilometers, the area is defined as several grids. Taking each grid point as the center and considering the terrain characteristics and ease of installation, each observation point is set within a radius of 0.5 kilometers, and micro-meteorological sensors are installed.

[0032] The micro-weather sensor collects data on the following parameters: irradiance, wind speed, wind direction, and temperature.

[0033] Data transmission method for micro-weather sensors: 5G.

[0034] Data collection frequency: 5 minutes.

[0035] S2, Extraction of Ideal Irradiance Curve

[0036] The ideal irradiance curve reflects the best irradiance achievable under clear weather conditions. Ideal irradiance is directly related to the solar altitude angle, but is also affected by the weakening effect of sunlight caused by moisture in the air.

[0037] In the same region, where meteorological conditions are similar, data can be processed together. It should be noted that the extracted ideal curves are derived with reference to deviations in sunrise and sunset times, and are extracted separately for different seasons.

[0038] 1) Group the observation points into grids with an east-west radius of 50 kilometers and a north-south radius of 10 kilometers;

[0039] 2) Using a 3-minute step, group the curves for different dates in the historical data according to the irradiation duration (sunset time - sunrise time);

[0040] 3) For each group, the daily irradiance curve with the longest irradiance duration is used as the standard.

[0041] Stretch each of the daily irradiance curves over time;

[0042] It is important to note that during the stretching process, the center point should remain unchanged, and the stretching should be symmetrical from left to right.

[0043] 4) In each daily irradiance curve, only the data for the time point when the weather is sunny is retained;

[0044] 5) Using 1 minute intervals as a scale, group the irradiance values ​​according to the goal of "shortest time interval", and remove outlier data at the highest rate of 5%.

[0045] 6) Extract data from each time point and perform linear interpolation on the missing data in 1-minute increments to obtain the ideal irradiance curve.

[0046] S3, Handling the impact of rain and snow weather

[0047] Rainfall and snowfall can have a significant impact on cloud formations, and it is necessary to analyze the changing trends of the impact of different weather types on irradiance under different irradiance conditions.

[0048] Weather types are categorized as follows: heavy rain, moderate rain, light rain, blizzard, heavy snow, moderate snow, and light snow. Wind force and direction primarily affect cloud distribution, an effect that will be analyzed in the following steps. Under these weather types, irradiance typically becomes 0 (or approaches 0). Changes in irradiance are mainly related to the prediction of these weather types.

[0049] Weather forecasting is influenced by multiple environmental factors such as water vapor transport and air uplift. It directly connects with meteorological forecast data to set the above-mentioned weather conditions and sets the corresponding irradiance to 0 based on the duration of the above-mentioned weather types.

[0050] After the aforementioned weather type ends, the clouds will disappear, and the irradiance will return to the ideal irradiance curve, which will be set according to the sunrise and sunset times and the predicted target time.

[0051] S4. Cloud Change Trend Prediction

[0052] This step deals with data patterns under weather conditions other than rain and snow.

[0053] Photovoltaic power generation resources are all installed and deployed at or near the ground; cloud cover at different altitudes at the same location will affect irradiance. Due to limitations, micro-weather data loggers can only be installed at ground level to collect meteorological data (temperature, wind speed, wind direction) near the ground.

[0054] Two trends will influence future irradiance at various locations. The overall calculation process is as follows:

[0055] 1) Wind force and direction near the ground: will have a direct impact on the movement of clouds, but cannot provide feedback on wind force and direction at different altitudes and outside the ground;

[0056] The function of clouds is to block sunlight to a certain extent. The blocking rate at a given moment...

[0057] This can be expressed using the following formula 1:

[0058]

[0059] Where R represents the current irradiance, R m This indicates the ideal irradiance at this point in time.

[0060] The function of wind is to blow clouds from one location to another; the change in cloud cover at a location includes the amount of cloud that enters and the amount of cloud that escapes during that period. When wind blows clouds from one location to another, it does not blow all the clouds over; the proportion that is transferred is represented by an attenuation coefficient.

[0061] By calculating the wind force and direction at different locations, the cloud movement positions at each point are determined after 1 minute, thus creating a cloud movement map.

[0062] Let the attenuation coefficient per kilometer be represented by λ1. This parameter is set to 0.5 by default. Subsequently, optimization is performed by calculating residuals from historical data over a certain period of time based on a certain step size.

[0063] Predict and assess the blocking rate at a point P. Based on the cloud transition map, determine the cloud origin P1 at point P0 one minute prior; for each P... i-1 Calculate the origin P of its cloud layer. i The trajectory sequence {P1, P2, ..., P} from which the cloud origin is obtained is obtained. n After n minutes of simulation, the predicted cloud cover is shown in Formula 2:

[0064]

[0065] Among them, B(P) i ) represents P i The current blocking rate.

[0066] 2) Local irradiance variation: This is a feedback of information on "wind force and direction at different cloud heights and not near the ground".

[0067] For a point P, when predicting its irradiance n minutes later, in order to continue the trend of irradiance variation and reflect unobservable information, an autoregressive model in time series is used for prediction. The blocking rate data takes values ​​between [0,1], and is less likely to be approximated closer to the boundary. Therefore, the blocking rate is transformed here, and the corresponding method is shown in Formula 3:

[0068]

[0069] Where N is a coefficient, with a default value of 10.

[0070] The transformed B' is predicted using an autoregressive model, and then solved in reverse using Equation 3 to obtain the predicted value B2.

[0071] 3) Finally, the overall blocking rate is calculated using the following formula 4:

[0072] B=B1×λ2+B2×(1-λ2) (4)

[0073] In the formula, the coefficient λ² ranges from (0,1), with a default value of 0.5 and a step size of 0.05. Analysis is performed based on prior values, and the effective coefficient is set by verifying the residuals. S5. Regional Full-Map Extrapolation and Irradiance Prediction

[0074] 1) Initialize the blocking rate of each point based on the observation point data.

[0075] The region is divided into grids with a grid length of 1 kilometer. For each grid point P, the k-nearest neighbor method is used to find the three nearest observation points {P1, P2, P3}, and the initial obstruction rate of the relevant grid points is calculated using Formula 5 below:

[0076]

[0077] In the formula, L(P,P) i ) represents P, P i The distance between two points.

[0078] The wind force and direction at each grid point are also based on the wind force and direction at the observation point, using k-nearest neighbors, and are evaluated according to the proportion of the inverse of the distance.

[0079] 2) Using the above steps, based on the previous time point t-1, obtain the blocking rate map of each grid point at the previous time point;

[0080] 3) Using step 4, deduce the blocking rate B4 at each position at time t;

[0081] 4) Finally, the overall blocking rate is calculated using the following formula 6:

[0082] B=B3×λ3+B4×(1-λ3) (6)

[0083] In the formula, the coefficient λ3 takes values ​​in the range of (0,1), with a default value of 0.5 and a step size of 0.05. The effective coefficient is set by verifying the residual.

[0084] 1. Full-map projection of future meteorological conditions based on changing trends

[0085] Using the results from step 5 above, with a grid length of 1 kilometer, set the wind force, wind direction, and obstruction rate for each grid point; then, using step 4, perform a full-map projection of future weather conditions within the next n minutes.

[0086] 2. Irradiance prediction based on ideal irradiance curve

[0087] Based on the weather conditions (blocking rate) at each time point within the next n minutes for each grid point, use Formula 7 below to calculate the irradiance of the corresponding grid point at the corresponding time point:

[0088] E(P,t)=B(P,t)×E m (P,t) (7)

[0089] In the formula, E(P,t) represents the predicted irradiance of location P at future time t, and B(P,t) represents the predicted blocking rate of location P at future time t. m (P,t) represents the calculated ideal irradiance value for location P at a future time t.

[0090] It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention.

Claims

1. A photovoltaic irradiance prediction method based on trend extrapolation, characterized in that, Includes the following steps: S1. Data Acquisition and Gridded Deployment: Within the target prediction area, multiple observation points are set at preset intervals, and micro-meteorological sensors are deployed at each observation point to collect micro-meteorological data including irradiance, wind force, and wind direction at a first preset frequency. S2. Constructing the ideal irradiance curve: Based on the historical irradiance data of the target prediction area, extract the ideal irradiance curves for different seasons under sunny weather conditions; S3. Identify and process the impact of rain and snow weather: Based on meteorological forecast data, identify the types of rain and snow weather, and set the irradiance forecast value for the corresponding weather type within the forecast period to zero; S4. Cloud Change Trend Inference: For non-rainy or snowy weather, based on the micro-meteorological data, combined with the cloud blocking rate model and spatiotemporal inference algorithm, the cloud blocking rate change trend of each location in the future period is predicted. The cloud obstruction rate B is determined by the formula... Calculate, where R represents the current irradiance and Rm represents the ideal irradiance at this point in time; S5. Regional Irradiance Prediction: By integrating the ideal irradiance curve with the predicted cloud blocking rate, the predicted irradiance values ​​for each location within the target prediction area in the future time period are calculated.

2. The method according to claim 1, characterized in that, The construction of the ideal irradiance curve in step S2 specifically includes: S21. Group the historical irradiance data according to geographical coordinates and season; S22. For each group, using the irradiance curve of the day with the longest irradiance duration as the benchmark, the irradiance curves of other days are time-stretched and aligned with the midday point as the center of symmetry. S23. Filter out data points with sunny weather from the aligned curves and remove outliers; S24. Interpolate the processed data to generate the ideal irradiance curve in minutes.

3. The method according to claim 1, characterized in that, In step S4, the cloud change trend inference includes cloud transfer inference based on near-surface wind force and direction and autoregressive inference based on local historical irradiance data. The cloud transfer extrapolation based on near-surface wind force and direction: Calculate the cloud origin trajectory sequence {P1,P2,...,P} based on wind force and direction. n }, and through formula Predict the blocking rate B1 at location P in the next n minutes, where λ1 is the cloud attenuation coefficient per kilometer, and B(P i ) represents position P i The current blocking rate; Autoregressive extrapolation based on local historical irradiance data: After logarithmically transforming the historical blocking rate data, the autoregressive model is used to predict the blocking rate B2 in the next n minutes.

4. The method according to claim 3, characterized in that, In step S4, the final position P after n minutes of the future has a comprehensive blocking rate B3 calculated by the formula B=B1×λ2+B2×(1-λ2), where λ2 is a weighting coefficient, the value of which is determined by optimizing the residual through historical data verification.

5. The method according to claim 1, characterized in that, Before step S5, the method also includes step S0, a full-map simulation of current meteorological conditions: S01. Divide the target prediction area into a dense grid with a grid length of 1 kilometer; S02. Based on the real-time obstruction rate of each observation point, the k-nearest neighbor method is used to find the three nearest observation points for each grid point, and the formula is used to... Calculate the initial blocking rate for each grid point, where L(P,P) i ) represents P, P i The distance between two points; S03. Starting from the initial blocking rate distribution map, execute step S4 to deduce the blocking rate distribution at future times.

6. The method according to claim 5, characterized in that, In step S03, the deduced blocking rate B4 at time t and the initial blocking rate B3 are fused using the formula B=B3×λ3+B4×(1-λ3), where λ3 is a weighting coefficient, the value of which is determined by verification residual optimization.

7. The method according to claim 1, characterized in that, In step S5, the formula E(P,t)=B(P,t)×E is used. m The irradiance prediction E(P,t) for location P at future time t is calculated, where B(P,t) represents the predicted blocking rate of location P at future time t, and E... m (P,t) represents the calculated ideal irradiance value for location P at a future time t.

8. The method according to claim 1, characterized in that, The first preset frequency is once every 5 minutes, the future time period is 2 hours, the preset spacing is 10 kilometers, and the observation points are deployed within a radius of 0.5 kilometers centered on the grid points.

Images