CN-120497977-B - Photovoltaic power generation intelligent scheduling method and device based on dynamic supply and demand prediction

Abstract

The invention relates to an intelligent photovoltaic power generation scheduling method and device based on dynamic supply and demand prediction. The photovoltaic power generation intelligent scheduling method based on dynamic supply and demand prediction comprises the steps of obtaining position information of a current cloud cluster, combining the position information with wind speed data, conducting gridding prediction through Kalman filtering to obtain a cloud cluster coverage state, conducting shielding calculation on the cloud cluster coverage state through a shielding probability function to obtain cloud cluster shielding probability, conducting joint calculation on the cloud cluster coverage state and the cloud cluster shielding probability through an illumination intensity attenuation model and a photoelectric power linear model to obtain a photovoltaic power generation power prediction value, conducting difference calculation on the photovoltaic power generation power prediction value and an electricity load prediction curve to obtain predicted supply and demand power, and conducting scheduling distribution on an electricity storage system according to the predicted supply and demand power to complete predictive power supply scheduling of electricity. The intelligent photovoltaic power generation scheduling method based on dynamic supply and demand prediction has the advantage of remarkably improving the real-time utilization rate of photovoltaic power under severe weather.

Inventors

• PAN ZHIJIE

Assignees

• 广东拓杰机电科技有限公司

Dates

Publication Date

20260508

Application Date

20250324

Claims (10)

1. 1. The intelligent photovoltaic power generation scheduling method based on dynamic supply and demand prediction is characterized by comprising the following steps of: S1, acquiring position information of a current cloud cluster, wherein the position information of the cloud cluster comprises a cloud cluster centroid position, a cloud cluster area, a cloud cluster speed state vector and a cloud layer thickness of the cloud cluster; S2, combining the position information of the current cloud cluster with wind speed data, and performing gridding prediction by adopting Kalman filtering to obtain a cloud cluster coverage state within a preset time, wherein the cloud cluster coverage state is the cloud cluster area in a prediction area; S3, carrying out shielding calculation on the cloud cover state in the preset time by adopting a shielding probability function to obtain cloud shielding probability in the preset time; S4, carrying out attenuation prediction on the cloud cover state and the cloud shielding probability within preset time by adopting an illumination intensity attenuation model to obtain a light intensity predicted value within the preset time; s5, converting the light intensity predicted value in the preset time by adopting a photoelectric power linear model to obtain a photovoltaic power generation power predicted value in the preset time; S6, performing difference calculation on the electricity load prediction curve in a preset time and the corresponding photovoltaic power generation power prediction value to obtain predicted supply and demand power in the preset time; And S7, scheduling and distributing the power storage system according to the predicted supply and demand power within the preset time to complete the predictive power supply scheduling of the power utilization.
2. 2. The intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand prediction according to claim 1, wherein the step S2 comprises the following sub-steps: S21, gridding a prediction area according to cloud cluster area and wind speed data in the current cloud cluster position information to obtain a prediction area set after grid division; The specific representation of the prediction area set after meshing is as follows: in the formula, Representing a set of meshed prediction areas, wherein the prediction areas represent illumination intensity change areas caused by projections generated by cloud cluster moving in the sky; representing the first of the current predicted areas Line 1 Grid of columns including location information of current cloud, and total number of grids being The specific calculation is as follows: in the formula, Representing the total area of the predicted region; Representing an upward rounding; Representing a maximum grid number constraint; representing the actual area of a single grid, the specific calculations of which are shown below: in the formula, Representing a meshing size factor; neglecting a threshold for cloud impact; representing the adaptive mesh resolution adjustment factor, the specific calculation is as follows: in the formula, A module value representing wind speed data; representing a reference wind speed, obtained by statistically weighted averaging of historical data; Representing the cloud area in the current cloud position information; an influence factor indicating a change in wind speed and area; is a cloud area occupation ratio item; s22, predicting the prediction area set after grid division by adopting a Kalman filter in combination with wind speed data according to the preset time to obtain a cloud cover state in the preset time.
3. 3. The intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand prediction according to claim 2, wherein the specific iterative expression of the kalman filter is as follows: First, the Line 1 The cloud state of the column grid is: in the formula, Is the first Line 1 Cloud state of column grid; And Representing a grid A local cloud centroid location in (a); And Represent the first Line 1 The speed state of the cloud of column grid; Representing the rate of change of the cloud area; For the first Line 1 A state equation for a column grid, expressed as follows: in the formula, Is a grid At the predicted time Cloud state of (2); Is a grid The state transition matrix of (2) is specifically expressed as follows: in the formula, Is the time step; An attenuation term indicating the rate of change of the area, Expressed in cloud diffusion time constant; The wind speed control item is specifically expressed as follows: in the formula, And The influence weight is used for expressing the wind speed on the cloud cluster speed; is process noise; Is an external control input, namely an observation variable; Indicating that the wind speed is at And The continuous velocity component in the direction is obtained by predicting by using LSTM in combination with historical data, and is specifically expressed as follows: in the formula, Historical wind speed data representing an initial measurement; wind speed data representing a current measurement; Representation of Predicted wind speed data for the moment; Memorizing a recurrent neural network for a long and short time of the predicted wind speed, and performing training fitting on the recurrent neural network by adopting historical data as a training data set to obtain the recurrent neural network; Representing a neighboring mesh versus a current mesh Is a weight matrix of influence of (1); an impact matrix representing the current grid by the neighboring grid is specifically expressed as follows: in the formula, Representation of State vectors of neighboring meshes of the moment; And For representing grids A corresponding cloud speed state; The range of influence is represented, and the specific calculation is as follows: in the formula, Representing a reference impact range; Representation of Cloud area change rate of the current grid at moment; is the area change rate sensitivity coefficient; Grid influence weights representing lateral and longitudinal offsets, which are specifically calculated as: in the formula, The Gaussian decay weight value is specifically expressed as follows: in the formula, Is a grid offset; And Representing the actual side length of the individual grid, i.e. ; For representing a domain grid An actual position offset relative to the current grid; representing a cloud displacement vector affected by wind speed; the specific calculation is expressed as follows: in the formula, Is the standard deviation of the reference; is the sensitivity coefficient of standard deviation; For the first Time grid The prediction of the kalman filter of cloud states of (a) is specifically expressed as follows: in the formula, Representing a grid Is at (1) Predicted cloud cluster state at time; Is a grid Is a prediction error covariance matrix of (a); Representing a grid At the position of A posterior error covariance matrix of the moment; Representing a grid Is a process noise covariance matrix; When predicting to obtain the first Time grid After cloud cluster state, carrying out Kalman filtering update, which is specifically expressed as follows: in the formula, Representing a grid A Kalman filter matrix of (2); Representing a grid Is a matrix of observations of (a); a covariance matrix representing observed noise; representing the first obtained by LSTM prediction The pseudo observed value of the moment is used for replacing real observed data, and is specifically expressed as follows: in the formula, Representing an initial measured cloud state; Representing the actual cloud cluster state of the current measurement; a recurrent neural network is memorized for predicting the long and short time of cloud cluster state; wherein, when the Kalman filtering update is completed within the preset time, for the first The specific calculation of cloud cover state in the global preset time of moment is expressed as follows: in the formula, For indicating the first The cloud cover state corresponding to the moment, namely Cloud area in the prediction area at any time; Representing a grid Is a part of the initial cloud area; Representing a grid Kalman filtering of (1) at And predicting the obtained area change rate at the moment.
4. 4. The intelligent photovoltaic power generation scheduling method based on dynamic supply and demand prediction according to claim 3, wherein the shielding probability function calculates the first The cloud occlusion probability at the moment is specifically expressed as follows: in the formula, Represent the first Cloud occlusion probability at moment; representing a historical occlusion probability term; Is the first The historical occlusion weight of the moment is specifically calculated as follows: in the formula, Represent the first The area change rate of cloud cover at the moment; representing historical impact adjustment coefficients; representing the coverage degree of the current cloud cluster on the prediction area; Represent the first The cloud diffusion probability gain at the moment is specifically calculated as follows: in the formula, Representing a diffusion-influencing-adjustment coefficient; Indicating the degree of change in saturation, which Is a saturation threshold; the remaining space representing the predicted area, i.e., the diffusion space.
5. 5. The intelligent photovoltaic power generation scheduling method based on dynamic supply and demand prediction according to claim 4, wherein for the first obtained by calculation of the illumination intensity decay model The specific calculation of the predicted value of the illumination intensity at the moment is shown as follows: in the formula, Indicating the final first A predicted value of the light intensity at the moment; representing a history compensation term, which The historical light intensity compensation weight coefficient is represented, and the concrete calculation is as follows: in the formula, Representing the historical data influence attenuation rate; Represent the first The rate of change of the probability of occlusion at the moment; representing a diffusion compensation term, the specific calculation of which is represented as follows: in the formula, A coverage area change rate threshold representing a cloud; No history compensation The predicted value of the illumination intensity at the moment is specifically expressed as follows: in the formula, Representing the current clear sky radiation intensity; Representing an effective optical path term, which The specific calculation of the cloud area attenuation coefficient is shown as follows: in the formula, Representing cloud type coefficients; Representing the cloud layer thickness; representing a cloud reference height; Indicating a light transmittance correction term, which The weight coefficient representing the cloud thickness is specifically calculated as follows: in the formula, Representing a base optical thickness coefficient; Representing a thickness sensitivity coefficient, and obtaining the thickness sensitivity coefficient through Mie scattering calculation; representing the equivalent path length correction coefficient, the specific calculation is as follows: in the formula, Represent the first The altitude of the sun at the moment.
6. 6. The intelligent photovoltaic power generation scheduling method based on dynamic supply and demand prediction according to claim 5, wherein the photovoltaic power linear model is used for conversion to obtain the first photovoltaic power generation scheduling method The specific representation of the photovoltaic power generation power predicted value at the moment is as follows: in the formula, Indicating the final first A photovoltaic power generation power prediction value at a moment; no history smoothing representation The predicted value of photovoltaic power generation power at the moment is specifically expressed as follows: in the formula, Representing the comprehensive efficiency coefficient of the photovoltaic module; Is the total effective photovoltaic unit of the photovoltaic power station; representing a photovoltaic power generation power loss term; Representing a history smoothing compensation term; the dynamic smoothing weight for the generated power is specifically calculated as follows: in the formula, Historical smooth adjustment coefficients for photovoltaic power generation power; Is the rate of change of the generated power.
7. 7. The intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand prediction according to claim 6, wherein the specific acquisition mode of the power load prediction curve in the preset time comprises the following steps: s61, sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence; S62A, carrying out statistical calculation on the electric load data sequence to obtain a statistical feature vector of the electric load data; S62B, carrying out wavelet decomposition on the electric load data sequence by adopting one-dimensional Haar wavelet to obtain high-frequency and low-frequency characteristic vectors of the electric load data; s63, splicing and fusing the statistical feature vector of the power load data and the high-frequency and low-frequency feature vector to obtain a fused feature vector; S64, predicting the fusion feature vector by adopting an electric load regression model to obtain an electric load prediction curve in a preset time; wherein, after obtaining the electricity load prediction curve in the preset time, for the first The concrete representation of the difference calculation of the time is as follows: in the formula, Represent the first Predicting supply and demand power at moment; representing the first of the electrical load prediction curves A predicted value of the power load at the moment; representing a power supply transmission loss coefficient; As a safety margin, the specific calculation is as follows: in the formula, Representing the fluctuation intensity coefficient; representing an uncertainty coefficient; means for representing high-frequency components in the high-frequency and low-frequency eigenvectors of the electrical load data; the standard deviation of the high-frequency component in the high-frequency and low-frequency feature vector of the electrical load data is represented.
8. 8. The intelligent photovoltaic power generation scheduling device based on dynamic supply and demand prediction is characterized by comprising a cloud cluster position information acquisition unit, a cloud cluster position prediction unit, a cloud cluster shielding probability calculation unit, a light intensity attenuation prediction unit, a light intensity-power conversion unit, a supply and demand power calculation unit and a predictive power supply scheduling unit; the cloud cluster position information acquisition unit is used for acquiring the position information of the current cloud cluster, wherein the position information of the cloud cluster comprises a cloud cluster centroid position, a cloud cluster area, a cloud cluster speed state vector and a cloud layer thickness of the cloud cluster; The cloud cluster position prediction unit is used for combining the position information of the current cloud cluster with wind speed data, and performing gridding prediction by adopting Kalman filtering to obtain a cloud cluster coverage state within preset time, wherein the cloud cluster coverage state is the cloud cluster area in a prediction area; The cloud cover probability calculation unit is used for carrying out shielding calculation on the cloud cover state in the preset time by adopting a shielding probability function to obtain the cloud cover probability in the preset time; The light intensity attenuation prediction unit is used for carrying out attenuation prediction on the cloud cover state and the cloud shielding probability within the preset time by adopting an illumination intensity attenuation model to obtain a light intensity predicted value within the preset time; The light intensity-power conversion unit is used for converting the light intensity predicted value in the preset time by adopting a photoelectric power linear model to obtain a photovoltaic power generation predicted value in the preset time; The supply and demand power calculation unit is used for carrying out difference calculation on the electricity load prediction curve in a preset time and the corresponding photovoltaic power generation power prediction value to obtain predicted supply and demand power in the preset time; And the predictive power supply scheduling unit is used for scheduling and distributing the power storage system according to the predicted power supply and demand in the preset time so as to complete the predictive power supply scheduling of the power consumption.
9. 9. The intelligent photovoltaic power generation scheduling apparatus based on dynamic supply and demand prediction according to claim 8, wherein the cloud cluster position prediction unit is further configured to perform the sub-steps of: S21, gridding a prediction area according to cloud cluster area and wind speed data in the current cloud cluster position information to obtain a prediction area set after grid division; The specific representation of the prediction area set after meshing is as follows: in the formula, Representing a set of meshed prediction areas, wherein the prediction areas represent illumination intensity change areas caused by projections generated by cloud cluster moving in the sky; representing the first of the current predicted areas Line 1 Grid of columns including location information of current cloud, and total number of grids being The specific calculation is as follows: in the formula, Representing the total area of the predicted region; Representing an upward rounding; Representing a maximum grid number constraint; representing the actual area of a single grid, the specific calculations of which are shown below: in the formula, Representing a meshing size factor; neglecting a threshold for cloud impact; representing the adaptive mesh resolution adjustment factor, the specific calculation is as follows: in the formula, A module value representing wind speed data; representing a reference wind speed, obtained by statistically weighted averaging of historical data; Representing the cloud area in the current cloud position information; an influence factor indicating a change in wind speed and area; is a cloud area occupation ratio item; s22, predicting a prediction area set after grid division by adopting a Kalman filter in combination with wind speed data according to preset time to obtain a cloud cover state in the preset time; the specific iterative expression of the Kalman filter is as follows: First, the Line 1 The cloud state of the column grid is: in the formula, Is the first Line 1 Cloud state of column grid; And Representing a grid A local cloud centroid location in (a); And Represent the first Line 1 The speed state of the cloud of column grid; Representing the rate of change of the cloud area; For the first Line 1 A state equation for a column grid, expressed as follows: in the formula, Is a grid At the predicted time Cloud state of (2); Is a grid The state transition matrix of (2) is specifically expressed as follows: in the formula, Is the time step; An attenuation term indicating the rate of change of the area, Expressed in cloud diffusion time constant; The wind speed control item is specifically expressed as follows: in the formula, And The influence weight is used for expressing the wind speed on the cloud cluster speed; is process noise; Is an external control input, namely an observation variable; Indicating that the wind speed is at And The continuous velocity component in the direction is obtained by predicting by using LSTM in combination with historical data, and is specifically expressed as follows: in the formula, Historical wind speed data representing an initial measurement; wind speed data representing a current measurement; Representation of Predicted wind speed data for the moment; Memorizing a recurrent neural network for a long and short time of the predicted wind speed, and performing training fitting on the recurrent neural network by adopting historical data as a training data set to obtain the recurrent neural network; Representing a neighboring mesh versus a current mesh Is a weight matrix of influence of (1); an impact matrix representing the current grid by the neighboring grid is specifically expressed as follows: in the formula, Representation of State vectors of neighboring meshes of the moment; And For representing grids A corresponding cloud speed state; The range of influence is represented, and the specific calculation is as follows: in the formula, Representing a reference impact range; Representation of Cloud area change rate of the current grid at moment; is the area change rate sensitivity coefficient; Grid influence weights representing lateral and longitudinal offsets, which are specifically calculated as: in the formula, The Gaussian decay weight value is specifically expressed as follows: in the formula, Is a grid offset; And Representing the actual side length of the individual grid, i.e. ; For representing a domain grid An actual position offset relative to the current grid; representing a cloud displacement vector affected by wind speed; the specific calculation is expressed as follows: in the formula, Is the standard deviation of the reference; is the sensitivity coefficient of standard deviation; For the first Time grid The prediction of the kalman filter of cloud states of (a) is specifically expressed as follows: in the formula, Representing a grid Is at (1) Predicted cloud cluster state at time; Is a grid Is a prediction error covariance matrix of (a); Representing a grid At the position of A posterior error covariance matrix of the moment; Representing a grid Is a process noise covariance matrix; When predicting to obtain the first Time grid After cloud cluster state, carrying out Kalman filtering update, which is specifically expressed as follows: in the formula, Representing a grid A Kalman filter matrix of (2); Representing a grid Is a matrix of observations of (a); a covariance matrix representing observed noise; representing the first obtained by LSTM prediction The pseudo observed value of the moment is used for replacing real observed data, and is specifically expressed as follows: in the formula, Representing an initial measured cloud state; Representing the actual cloud cluster state of the current measurement; a recurrent neural network is memorized for predicting the long and short time of cloud cluster state; wherein, when the Kalman filtering update is completed within the preset time, for the first The specific calculation of cloud cover state in the global preset time of moment is expressed as follows: in the formula, For indicating the first The cloud cover state corresponding to the moment, namely Cloud area in the prediction area at any time; Representing a grid Is a part of the initial cloud area; Representing a grid Kalman filtering of (1) at Predicting the obtained area change rate at the moment; Wherein the occlusion probability function calculates the first The cloud occlusion probability at the moment is specifically expressed as follows: in the formula, Represent the first Cloud occlusion probability at moment; representing a historical occlusion probability term; Is the first The historical occlusion weight of the moment is specifically calculated as follows: in the formula, Represent the first The area change rate of cloud cover at the moment; representing historical impact adjustment coefficients; representing the coverage degree of the current cloud cluster on the prediction area; Represent the first The cloud diffusion probability gain at the moment is specifically calculated as follows: in the formula, Representing a diffusion-influencing-adjustment coefficient; Indicating the degree of change in saturation, which Is a saturation threshold; a residual space representing the predicted area, i.e., a diffusible space; wherein, for the first obtained by calculation of the illumination intensity decay model The specific calculation of the predicted value of the illumination intensity at the moment is shown as follows: in the formula, Indicating the final first A predicted value of the light intensity at the moment; representing a history compensation term, which The historical light intensity compensation weight coefficient is represented, and the concrete calculation is as follows: in the formula, Representing the historical data influence attenuation rate; Represent the first The rate of change of the probability of occlusion at the moment; representing a diffusion compensation term, the specific calculation of which is represented as follows: in the formula, A coverage area change rate threshold representing a cloud; No history compensation The predicted value of the illumination intensity at the moment is specifically expressed as follows: in the formula, Representing the current clear sky radiation intensity; Representing an effective optical path term, which The specific calculation of the cloud area attenuation coefficient is shown as follows: in the formula, Representing cloud type coefficients; Representing the cloud layer thickness; representing a cloud reference height; Indicating a light transmittance correction term, which The weight coefficient representing the cloud thickness is specifically calculated as follows: in the formula, Representing a base optical thickness coefficient; Representing a thickness sensitivity coefficient, and obtaining the thickness sensitivity coefficient through Mie scattering calculation; representing the equivalent path length correction coefficient, the specific calculation is as follows: in the formula, Represent the first The altitude of the sun at the moment; wherein the photoelectric power linear model is adopted for conversion to obtain the first The specific representation of the photovoltaic power generation power predicted value at the moment is as follows: in the formula, Indicating the final first A photovoltaic power generation power prediction value at a moment; no history smoothing representation The predicted value of photovoltaic power generation power at the moment is specifically expressed as follows: in the formula, Representing the comprehensive efficiency coefficient of the photovoltaic module; Is the total effective photovoltaic unit of the photovoltaic power station; representing a photovoltaic power generation power loss term; Representing a history smoothing compensation term; the dynamic smoothing weight for the generated power is specifically calculated as follows: in the formula, Historical smooth adjustment coefficients for photovoltaic power generation power; Is the rate of change of the generated power.
10. 10. The intelligent photovoltaic power generation scheduling device based on dynamic supply and demand prediction according to claim 8, wherein the specific acquisition mode of the power load prediction curve within the preset time comprises the following steps: s61, sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence; S62A, carrying out statistical calculation on the electric load data sequence to obtain a statistical feature vector of the electric load data; S62B, carrying out wavelet decomposition on the electric load data sequence by adopting one-dimensional Haar wavelet to obtain high-frequency and low-frequency characteristic vectors of the electric load data; s63, splicing and fusing the statistical feature vector of the power load data and the high-frequency and low-frequency feature vector to obtain a fused feature vector; S64, predicting the fusion feature vector by adopting an electric load regression model to obtain an electric load prediction curve in a preset time; wherein, after obtaining the electricity load prediction curve in the preset time, for the first The concrete representation of the difference calculation of the time is as follows: in the formula, Represent the first Predicting supply and demand power at moment; representing the first of the electrical load prediction curves A predicted value of the power load at the moment; representing a power supply transmission loss coefficient; As a safety margin, the specific calculation is as follows: in the formula, Representing the fluctuation intensity coefficient; representing an uncertainty coefficient; means for representing high-frequency components in the high-frequency and low-frequency eigenvectors of the electrical load data; the standard deviation of the high-frequency component in the high-frequency and low-frequency feature vector of the electrical load data is represented.

Description

Photovoltaic power generation intelligent scheduling method and device based on dynamic supply and demand prediction Technical Field The invention relates to the field of supply and demand prediction of power supply or distribution, in particular to an intelligent photovoltaic power generation scheduling method and device based on dynamic supply and demand prediction. Background With the attention of modern society to renewable energy sources, photovoltaic power generation is a clean and efficient renewable energy source, and has become an energy technology widely applied and deployed on a large scale. Conventional photovoltaic power generation stations are usually built in arid and rainless areas such as plain, desert, mountain land and the like, and the photovoltaic power generation system often needs to occupy a larger area of land. Therefore, coastal areas, particularly offshore photovoltaic projects, are becoming a new renewable energy development direction, aiming at maximizing the utilization of solar energy resources in open areas on the sea. However, although offshore photovoltaic is a novel resource development mode, in practical application, photovoltaic power generation in coastal areas is often severely affected by meteorological factors, especially in typhoon seasons and marine climates, meteorological changes such as cloud layer thickness, cloud cluster movement tracks and the like are easy to become extremely complex and difficult to predict, so that attenuation of illumination intensity is highly nonlinear and even severe fluctuation is likely to occur, and the fluctuation of illumination intensity directly affects power generation power of a photovoltaic power station, so that prediction of power generation amount becomes difficult and unstable. In practical application, especially in micro-grids of photovoltaic power stations, the real-time power generation power of the photovoltaic power stations is usually used as a main energy supply source preferentially, so that loss in energy conversion and transmission processes is reduced as much as possible, when the real-time power generation power of the photovoltaic power stations is insufficient to meet the current power utilization load, an electric power storage system is required to be called for electric quantity compensation, and when the electric power storage capacity is insufficient, an external or main power grid is finally started to supply power by a dispatching system, so that under extreme climate conditions such as marine climate or typhoon, the photovoltaic power generation is greatly fluctuated due to the influence of meteorological factors such as rapid cloud cluster change, severe fluctuation of wind speed and the like, meanwhile, the phenomenon of sudden change of power utilization requirements can occur in industrial production in coastal areas, and the double uncertainty of supply and demand greatly improves the dispatching difficulty of the electric power system, thereby influencing the stability and the power supply safety. In the prior art, the illumination intensity and the electricity load are usually calculated by adopting a prediction model based on a relatively stable climate, and the electricity load is balanced by combining a dispatching storage battery system, however, the marine climate and the typhoon climate have relatively strong uncertainty, so that the existing prediction model of the illumination intensity is easy to generate relatively large errors under the condition. Especially, in a short time when typhoons occur, the movement speed and the change direction of the cloud layer are unpredictable, so that the traditional method is difficult to cope with the severe change, delay and error of power supply scheduling are caused, and unstable power supply and potential safety hazards of power supply are caused. Accordingly, the prior art suffers from insufficient accuracy in dealing with predictive power supply scheduling in marine and typhoon climates. Disclosure of Invention Based on the above, the invention aims to provide a photovoltaic power generation intelligent scheduling method based on dynamic supply and demand prediction. A photovoltaic power generation intelligent scheduling method based on dynamic supply and demand prediction comprises the following steps: S1, acquiring position information of a current cloud cluster, wherein the position information of the cloud cluster comprises a cloud cluster centroid position, a cloud cluster area, a cloud cluster speed state vector and a cloud layer thickness of the cloud cluster; S2, combining the position information of the current cloud cluster with wind speed data, and performing grid prediction by adopting Kalman filtering to obtain a cloud cluster coverage state within a preset time; S3, carrying out shielding calculation on the cloud cover state in the preset time by adopting a shielding probability function to obtain cloud shielding probability in