A photovoltaic power generation intelligent scheduling method and device based on dynamic supply and demand prediction

Abstract

The application relates to a photovoltaic power generation intelligent scheduling method and device based on dynamic supply-demand prediction. The photovoltaic power generation intelligent scheduling method based on dynamic supply-demand prediction comprises the following steps: acquiring position information of a current cloud cluster, combining the position information with wind speed data, performing grid prediction by using Kalman filtering to obtain a cloud cluster coverage state; performing shading calculation on the cloud cluster coverage state by using a shading probability function to obtain a cloud cluster shading probability; and jointly calculating the cloud cluster coverage state and the cloud cluster shading probability by using a light intensity attenuation model and a photoelectric power linear model to obtain a photovoltaic power generation power prediction value, and performing difference calculation on the photovoltaic power generation power prediction value and an electricity load prediction curve to obtain a predicted supply-demand power; and scheduling and distributing an electricity storage system according to the predicted supply-demand power to complete predictive power supply scheduling for electricity. The photovoltaic power generation intelligent scheduling method based on dynamic supply-demand prediction has the advantage that the real-time utilization rate of photovoltaic electric energy can be significantly improved under severe weather.

Description

Technical Field

[0001] This invention relates to the field of power supply or distribution forecasting, and in particular to a method and apparatus for intelligent dispatching of photovoltaic power generation based on dynamic supply and demand forecasting. Background Technology

[0002] With increasing attention to renewable energy in modern society, photovoltaic (PV) power generation, as a clean and efficient renewable energy source, has become a widely used and large-scale deployed energy technology. Traditional PV power plants are typically built in arid, sunny areas such as plains, deserts, and mountains, but PV systems often require a large land area. Therefore, coastal areas, especially offshore PV projects, are gradually becoming a new direction for renewable energy development, aiming to maximize the utilization of solar energy resources in open offshore areas.

[0003] However, although offshore photovoltaics is a relatively new way of resource development, in practical applications, photovoltaic power generation in coastal areas is often severely affected by meteorological factors, especially during typhoon season and marine climate. This can easily lead to extremely complex and unpredictable meteorological changes such as cloud thickness and cloud movement trajectories, resulting in highly nonlinear attenuation of sunlight intensity, and even drastic fluctuations. These fluctuations in sunlight intensity directly affect the power generation of photovoltaic power plants, making the prediction of power generation difficult and unstable.

[0004] In practical applications, especially in microgrids of photovoltaic power plants, the real-time power generation of the photovoltaic power plant is usually prioritized as the main energy source to minimize losses during energy conversion and transmission. When the real-time power generation of the photovoltaic power plant is insufficient to meet the current electricity load, the energy storage system needs to be called in for power compensation. Only when the energy storage capacity is insufficient will the dispatch system finally activate the external or main power grid for power supply. Therefore, under extreme weather conditions such as maritime climates or typhoons, due to the influence of meteorological factors such as rapid changes in cloud formations and drastic fluctuations in wind speed, not only does photovoltaic power generation fluctuate significantly, but industrial production in coastal areas may also experience sudden changes in electricity demand. This dual uncertainty in supply and demand greatly increases the difficulty of power system dispatch, thereby affecting the stability and security of power supply.

[0005] Existing technologies typically employ predictive models based on relatively stable climates to calculate solar radiation intensity and electricity load, combining this with the scheduling of battery systems to balance the load. However, marine and typhoon climates are highly uncertain, making existing solar radiation intensity prediction models prone to significant errors in such situations. Especially during the short period of a typhoon, the unpredictable speed and direction of cloud movement make it difficult for traditional methods to cope with such drastic changes, leading to delays and errors in power dispatching, ultimately resulting in power instability and potential safety hazards. Therefore, existing technologies face insufficient accuracy in predictive power dispatching under marine and typhoon climates. Summary of the Invention

[0006] Therefore, the purpose of this invention is to provide a method for intelligent scheduling of photovoltaic power generation based on dynamic supply and demand forecasting.

[0007] A method for intelligent dispatching of photovoltaic power generation based on dynamic supply and demand forecasting includes the following steps:

[0008] S1. Obtain the current cloud cluster's location information; wherein, the cloud cluster's location information includes the cloud cluster's centroid location, cloud cluster area, cloud cluster's velocity state vector, and cloud layer thickness;

[0009] S2. Combine the current cloud location information with wind speed data, and use Kalman filtering to perform gridded prediction to obtain the cloud coverage status within a preset time.

[0010] S3. Use an occlusion probability function to calculate the cloud coverage state within a preset time period and obtain the cloud occlusion probability within the preset time period.

[0011] S4. Use a light intensity attenuation model to attenuate and predict the cloud coverage status and cloud occlusion probability within a preset time period, and obtain the light intensity prediction value within the preset time period.

[0012] S5. A photoelectric power linear model is used to convert the predicted light intensity value within a preset time period to obtain the predicted photovoltaic power generation value within the preset time period.

[0013] S6. Calculate the difference between the predicted electricity load curve and the corresponding predicted photovoltaic power generation within a preset time period to obtain the predicted supply and demand power within the preset time period.

[0014] S7. Based on the predicted power supply and demand within a preset time period, the energy storage system is scheduled and allocated to complete the predictive power supply scheduling for power consumption.

[0015] The photovoltaic power generation intelligent scheduling method based on dynamic supply and demand forecasting described in this invention, compared with the prior art, predicts the cloud movement trajectory in the future period and combines the light intensity attenuation model and the photovoltaic power linear model for conversion calculation to obtain the photovoltaic power generation in the future period. This enables the energy storage system to be adjusted in advance, improves the real-time utilization rate of photovoltaic power, and effectively improves the response speed of the scheduling strategy to avoid power shortages or waste caused by sudden weather changes.

[0016] Meanwhile, the Kalman filter of this invention uses wind speed prediction with time-forward (LSTM prediction) as a control term to perform gridded prediction of the current cloud location information, thereby significantly improving the accuracy of cloud change prediction under sudden climate change.

[0017] Further, step S2 includes the following sub-steps:

[0018] S21. Based on the cloud area and wind speed data in the current cloud location information, the prediction area is gridded to obtain a set of prediction areas after gridding.

[0019] The specific representation of the predicted region set after gridding is as follows:

[0020] G = {G i,j |i,j∈[1,M]}

[0021] In the formula, G represents the set of prediction regions after gridding, where the prediction region refers to the area of ​​light intensity variation caused by the projection of cloud movement across the sky; G i,j This represents the grid cell in the i-th row and j-th column of the current prediction region, including the location information of the current cloud cluster; the total number of grid cells is M×M, and its specific calculation is as follows:

[0022]

[0023] In the formula, Area represents the total area of ​​the predicted region; Indicates rounding up; M max Indicates the maximum number of grid cells constraint; Size grid The actual area of ​​a single grid cell is represented by the following calculation:

[0024]

[0025] In the formula, Resolution represents the mesh size factor; The threshold for ignoring the impact of cloud clusters;

[0026] λ res The adaptive grid resolution adjustment factor is calculated as follows:

[0027]

[0028] In the formula, ||u|| represents the modulus of the wind speed data; u base The reference wind speed is obtained through a statistically weighted average of historical data; A cloud This represents the cloud area in the current cloud location information; β represents the influencing factor of wind speed and area change. This refers to the percentage of cloud area.

[0029] S22. Based on the preset time, a Kalman filter is used in conjunction with wind speed data to predict the set of prediction areas after grid division, so as to obtain the cloud coverage status within the preset time.

[0030] Accordingly, this invention improves the accuracy and robustness of cloud trajectory prediction under sudden weather changes by adopting an adaptive grid partitioning strategy based on wind speed data and cloud area, while ensuring computational efficiency.

[0031] Furthermore, the specific iterative expression of the Kalman filter is as follows:

[0032] The cloud state of the grid in the i-th row and j-th column is:

[0033]

[0034] In the formula, x cloud,ij The cloud state is represented by the grid in the i-th row and j-th column; x c,ij and y c,ij Represents grid G ij The location of the local cloud centroid; v x and v y This represents the velocity state of the cloud cluster in the i-th row and j-th column. This represents the rate of change of the cloud area;

[0035] The state equation for the grid in the i-th row and j-th column is expressed as follows:

[0036]

[0037] In the formula, For grid G ij The cloud state at the predicted time k; F ij For grid G ij The state transition matrix is ​​specifically represented as follows:

[0038]

[0039] In the formula, Δt is the time step; The attenuation term represents the rate of change of area, and τ represents the cloud diffusion time constant;

[0040] B represents the wind speed control item, which is specifically expressed as follows:

[0041]

[0042] In the formula, β x and β y Used to represent the weight of the influence of wind speed on cloud velocity; For process noise; u (k) External control input, i.e., observed variables;

[0043] This represents the continuous velocity components of wind speed in the x and y directions. Specifically, it is obtained by using LSTM combined with historical data for prediction, and its specific representation is as follows:

[0044]

[0045] In the formula, u (init) This represents the historical wind speed data from the initial measurement; u (0) This indicates the currently measured wind speed data; u (k-1) This represents the predicted wind speed data at time k-1; LSTM 风速 To obtain the long short-term memory recurrent neural network for predicting wind speed, historical data was used as the training dataset to train and fit the network.

[0046] A ij Indicates the relationship between neighboring grids and the current grid G. ij The influence weight matrix; The matrix representing the influence of neighboring grids on the current grid is as follows:

[0047]

[0048] In the formula, v represents the state vector of the neighboring grid at time k+1; x,(i+m,j+n) and v y,(i+m,j+n) Used to represent grid G (i+m,j+n) The corresponding cloud velocity state; R represents the area of ​​influence, which is calculated as follows:

[0049]

[0050] In the formula, base represents the baseline influence range; γ represents the rate of change of cloud area in the current grid at time k-1; γ is the sensitivity coefficient for the rate of change of area.

[0051] α mn The weights representing the influence of grid offsets in the horizontal and vertical directions are calculated as follows:

[0052]

[0053] In the formula, distance mn The Gaussian decay weight value is represented as follows:

[0054]

[0055] In the formula, [m,n] represents the grid offset; Δx and Δy represent the actual side length of a single grid cell, i.e.

[0056] [m·Δx,n·Δt] is used to represent the actual position offset of the neighborhood grid [i+m,i+n] relative to the current grid;

[0057] σ represents the cloud displacement vector affected by wind speed; mn The standard deviation is calculated as follows:

[0058]

[0059] In the formula, σ base δ is the baseline standard deviation; δ is the standard deviation sensitivity coefficient.

[0060] For the grid G ​​at time k+1 ij The Kalman filter prediction of the cloud state is specifically represented as follows:

[0061]

[0062] In the formula, Represents grid G ij The predicted cloud state at time k+1; For grid G ij The prediction error covariance matrix; Represents grid G ij The posterior error covariance matrix at time k; Q ij Represents grid G ij The process noise covariance matrix;

[0063] When the prediction obtains the grid G ​​at time k+1 ij After determining the cloud state, a Kalman filter update is performed, as shown below:

[0064]

[0065] In the formula, Represents grid G ij Kalman filter matrix; H ij Represents grid G ij The observation matrix; R represents the covariance matrix of the observation noise; This represents the pseudo-observation value at time k+1 obtained through LSTM prediction, used to replace the real observation data. Its specific representation is as follows:

[0066]

[0067] In the formula, This indicates the initial cloud state. Indicates the actual cloud state currently being measured; LSTM 云团 A long short-term memory recurrent neural network for predicting cloud states;

[0068] Wherein, after the Kalman filter update within the preset time period is completed, the specific calculation representation of the global cloud coverage state within the preset time period at time k is as follows:

[0069]

[0070] In the formula, Cover (k) Used to represent the cloud cover state at time k, that is, the cloud area in the prediction region at time k; Represents grid G ij The initial cloud area; Represents grid G ij The Kalman filter predicts the rate of change of area at time r.

[0071] Accordingly, this invention uses LSTM to predict wind speed to enhance the long-term prediction accuracy of Kalman filtering for cloud movement trajectory, and introduces a dynamic weight coupling mechanism of neighboring grids to enhance the dynamic correlation between spatial distance, wind speed direction and area change rate, thereby ensuring that the predicted cloud movement trajectory has local consistency.

[0072] Furthermore, by generating pseudo-observations using LSTM, error compensation for long-term predictions is further improved, thereby enhancing the long-term prediction capability of the Kalman filter.

[0073] Furthermore, the specific representation of the cloud occlusion probability calculation at time k using the occlusion probability function is as follows:

[0074]

[0075] In the formula, This represents the probability of cloud cover at time k. This represents the historical occlusion probability term; Let be the historical occlusion weight at time k, and its specific calculation is as follows:

[0076]

[0077] In the formula, β represents the rate of change of the cloud cover area at time k; cover Indicates the historical impact adjustment coefficient;

[0078] This indicates the degree of coverage of the current cloud cluster over the predicted area; Let represent the cloud diffusion probability gain at time k, and its specific calculation is as follows:

[0079]

[0080] In the formula, γ diffusion Indicates the diffusion effect adjustment coefficient; The θ represents the degree of influence of the rate of change of saturation, and is the saturation threshold. This represents the remaining space of the predicted area, i.e., the space for expansion.

[0081] This invention employs an occlusion probability function to smooth the continuity of cloud changes, and combines historical occlusion weights to dynamically adjust the calculation method of occlusion probability, enabling it to better adapt to cloud states changing at different speeds.

[0082] Furthermore, by calculating cloud coverage rate to ensure that the occlusion probability is consistent with the actual cloud area, and by introducing a diffusion probability gain term to optimize the gradual change process of cloud boundary, the prediction accuracy of the impact of cloud movement on the degree of illumination is significantly improved, especially the predictive adaptability and accuracy of the degree of illumination impact under drastically changing climates such as typhoons and thunderstorms.

[0083] Furthermore, the specific calculation of the predicted light intensity value at time k obtained through the light intensity attenuation model is as follows:

[0084]

[0085] In the formula, This represents the final predicted light intensity value at time k. This represents the historical compensation item, whose w hist The historical light intensity compensation weighting coefficient is calculated as follows:

[0086]

[0087] In the formula, λ hist This indicates the impact of historical data on the decay rate; This represents the rate of change of the occlusion probability at time k; The diffusion compensation term is represented, and its specific calculation is as follows:

[0088]

[0089] In the formula, θ vIndicates the threshold for the rate of change of cloud cover area; I (k) The predicted illumination intensity at time k without historical compensation is represented as follows:

[0090]

[0091] In the formula, I0 represents the current clear-sky radiation intensity; This represents the effective optical path length term, where w area The cloud area attenuation coefficient is calculated as follows:

[0092]

[0093] In the formula, α tyoe Represents the cloud type coefficient; h cloud Indicates cloud thickness; h ref Indicates the reference height of the cloud cluster;

[0094] This represents the transmittance correction term, where μ represents the weighting coefficient for cloud thickness. Its specific calculation is as follows:

[0095] μ=μ0·(1-exp(-δ μ ·h cloud ))

[0096] In the formula, μ0 represents the basic optical thickness coefficient; δ μ The thickness sensitivity coefficient is obtained through Mie scattering calculations.

[0097] d eff The equivalent path length correction factor is calculated as follows:

[0098]

[0099] In the formula, This represents the solar altitude angle at time k.

[0100] Accordingly, this invention combines the predicted light intensity value with a historical compensation term to dynamically smooth changes in light intensity, thereby improving the continuity and accuracy of light intensity prediction. At the same time, it corrects the boundary effect of rapidly moving clouds by using a diffusion compensation term to avoid abnormal fluctuations in light intensity caused by sudden changes in cloud formations, thus improving the accuracy of light intensity prediction.

[0101] In addition, the light intensity prediction is dynamically adjusted by a transmittance correction term based on cloud thickness, solar altitude angle, and optical scattering characteristics, so that the prediction value is more consistent with the attenuation of the actual physical process.

[0102] Furthermore, the photovoltaic power linear model is used for conversion to obtain the specific representation of the photovoltaic power generation prediction value at time k, as follows:

[0103]

[0104] In the formula, This represents the final predicted photovoltaic power generation at time k. The predicted photovoltaic power generation at time k without historical smoothing is represented as follows:

[0105]

[0106] In the formula, η pv Indicates the overall efficiency coefficient of a photovoltaic module; A pv The total effective photovoltaic units of a photovoltaic power station; ∈ loss This represents the power loss term in photovoltaic power generation; Indicates the historical smoothing compensation term; w pv The dynamic smoothing weight for power generation is calculated as follows:

[0107]

[0108] In the formula, λ pv This is the historical smoothing adjustment coefficient for photovoltaic power generation. This represents the rate of change in power generation.

[0109] Accordingly, this invention introduces a dynamic smoothing mechanism based on historical data during the photovoltaic power conversion process. This mechanism uses a historical smoothing compensation term to integrate the current light intensity prediction with the historical power generation status, thereby achieving effective modeling of the lag characteristics of photovoltaic power generation. This makes the photovoltaic power generation prediction more closely resemble the actual response curve and improves the physical consistency of the prediction.

[0110] Furthermore, the specific method for obtaining the electricity load prediction curve within the preset time period includes the following steps:

[0111] S61. Sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence;

[0112] S62A. Perform statistical calculations on the electricity load data sequence to obtain the statistical feature vector of the electricity load data;

[0113] S62B: One-dimensional Haar wavelet is used to perform wavelet decomposition on the power load data sequence to obtain the high and low frequency feature vectors of the power load data.

[0114] S63. The statistical feature vector and high- and low-frequency feature vectors of the electricity load data are spliced ​​and fused to obtain a fused feature vector;

[0115] S64. Use the electricity load regression model to predict the fused feature vector and obtain the electricity load prediction curve within a preset time period.

[0116] Once the electricity load prediction curve for a preset time period is obtained, the specific representation of the difference calculation at time k is as follows:

[0117]

[0118] In the formula, This represents the predicted supply and demand power at time k; η represents the predicted electricity load at time k in the electricity load forecast curve. gird δ represents the power transmission loss coefficient. safe To provide a safety margin, the specific calculation is expressed as follows:

[0119] δ safe =λ safe ·μ(wave high )+γ safe ·σ(wave high )

[0120] In the formula, λ safe Indicates the wave intensity coefficient; γ safe μ represents the uncertainty coefficient; high σ(wave) represents the mean of the high-frequency components in the high- and low-frequency characteristic vectors of the electricity load data; high ) represents the standard deviation of the high-frequency component in the high- and low-frequency feature vectors of the electricity load data.

[0121] Accordingly, this invention dynamically samples historical electricity load data through a sliding window and integrates statistical feature extraction and Haar wavelet decomposition methods to simultaneously capture long-term trends and short-term abrupt changes in the load curve, thereby achieving highly sensitive identification and accurate prediction of load surges or drops, significantly improving the model's prediction stability and accuracy.

[0122] Furthermore, in the process of calculating the power supply and demand difference, this invention introduces a safety margin compensation mechanism based on high-frequency fluctuation characteristics. By using the mean and standard deviation of high-frequency components to quantitatively model the load fluctuation intensity, and dynamically adjusting the safety margin accordingly, the system's adaptive ability to sudden power consumption disturbances is enhanced, thereby effectively suppressing the supply and demand imbalance and power dispatch fluctuations caused by prediction errors.

[0123] A photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting includes a cloud cluster location information acquisition unit, a cloud cluster location prediction unit, a cloud cluster shading 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 dispatching unit.

[0124] The cloud cluster location information acquisition unit is used to acquire the current cloud cluster location information; wherein, the cloud cluster location information includes the cloud cluster centroid location, cloud cluster area, cloud cluster velocity state vector, and cloud layer thickness;

[0125] The cloud cluster location prediction unit is used to combine the current cloud cluster location information with wind speed data, and use Kalman filtering to perform gridded prediction to obtain the cloud cluster coverage status within a preset time.

[0126] The cloud occlusion probability calculation unit is used to calculate the cloud coverage state within a preset time using an occlusion probability function to obtain the cloud occlusion probability within the preset time.

[0127] The light intensity attenuation prediction unit is used to use a light intensity attenuation model to predict the cloud coverage state and cloud occlusion probability within a preset time, and obtain the light intensity prediction value within the preset time.

[0128] The light intensity-to-power conversion unit is used to convert the light intensity prediction value within a preset time using a photoelectric power linear model to obtain the photovoltaic power generation prediction value within the preset time.

[0129] The power supply and demand calculation unit is used to calculate the difference between the electricity load prediction curve and the corresponding photovoltaic power generation prediction value within a preset time period to obtain the predicted power supply and demand within the preset time period.

[0130] The predictive power supply scheduling unit is used to schedule and allocate the energy storage system according to the predicted power supply and demand within a preset time period, thereby completing the predictive power supply scheduling for electricity consumption.

[0131] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0132] Figure 1 This is a simplified structural diagram of the photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting described in this invention.

[0133] Figure 2 This is a simplified flowchart illustrating the intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand forecasting described in this invention.

[0134] Figure 3 This is a simplified diagram of the prediction area. Detailed Implementation

[0135] To address the insufficient accuracy of existing technologies in predictive power dispatching under maritime and typhoon conditions, this invention acquires the current cloud cluster location information and combines it with wind speed data. A Kalman filter algorithm is used to perform gridded prediction of the cloud clusters, obtaining the cloud cluster coverage state within a preset time period. An occlusion probability function is then used to calculate the occlusion probability of the cloud clusters within the preset time period. Next, based on the occlusion probability and cloud state, a light intensity attenuation model is used to predict the light intensity, thus obtaining the predicted light intensity value within the preset time period. This light intensity prediction value is then converted into a photovoltaic power generation prediction value using a linear photovoltaic power model. Finally, the difference between the predicted photovoltaic power generation value and the predicted electricity load value is calculated to obtain the predicted supply and demand power. Based on this predicted supply and demand power, the energy storage system is dispatched and allocated, completing the predictive power dispatching.

[0136] Accordingly, this invention calculates the predicted value of photovoltaic power generation based on weather forecasts and calculates it together with the predicted value of electricity load to accurately identify future power gaps or surpluses, and to carry out energy storage and power supply scheduling planning in advance, thereby improving the intelligent scheduling capability of local energy or microgrids under extreme weather conditions and avoiding energy waste or power shortages.

[0137] Based on the above design, this invention proposes a photovoltaic power generation intelligent scheduling method based on dynamic supply and demand forecasting, and based on this method, proposes a photovoltaic power generation intelligent scheduling device based on dynamic supply and demand forecasting.

[0138] Please also refer to Figure 1 and Figure 2 A simplified structural diagram of the photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting described in this invention. Figure 2 This is a simplified flowchart illustrating the intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand forecasting described in this invention.

[0139] The photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting includes a cloud cluster location information acquisition unit 1, a cloud cluster location prediction unit 2, a cloud cluster shading probability calculation unit 3, a light intensity attenuation prediction unit 4, a light intensity-power conversion unit 5, a supply and demand power calculation unit 6, and a predictive power supply dispatching unit 7.

[0140] The cloud cluster location information acquisition unit 1 is used to perform step S1: acquire the current cloud cluster location information.

[0141] Specifically, the location information of the cloud cluster includes the location of the cloud cluster's centroid, the cloud cluster's area, the cloud cluster's velocity state vector, and the cloud layer thickness.

[0142] The cloud centroid position is used to locate the core position of the current cloud. The reflected wave data of the cloud is obtained by meteorological radar, and the cloud outline is extracted from the reflected wave data. The centroid position is calculated from the cloud outline.

[0143] The cloud area is converted into cloud area based on the intensity of reflected waves from the cloud region in the weather radar image;

[0144] The cloud velocity vector is obtained by acquiring the radial velocity of the cloud using weather radar and combining it with multi-view data to calculate the horizontal and vertical movement velocities of the cloud.

[0145] The thickness of the cloud cluster is obtained by measuring the intensity and propagation time of the echo signal using a weather radar. It should be noted that the weather radar has a built-in cloud cluster positioning and tracking system, which can directly obtain the current location information parameters of the cloud cluster. This is not the subject of this invention and will not be described in detail here.

[0146] The cloud cluster location prediction unit 2 is used to perform step S2: combining the current cloud cluster location information with wind speed data, using Kalman filtering for gridded prediction, and obtaining the cloud cluster coverage status within a preset time period.

[0147] Specifically, the prediction area is gridded based on the cloud area and wind speed data from the current cloud location information, resulting in a set of gridded prediction areas, as shown below:

[0148] G = {G i,j |i,j∈[1,M]}

[0149] In the formula, G represents the set of prediction regions after gridding. The prediction region refers to the area of ​​change in light intensity caused by the projection of cloud movement across the sky. This change in the sky area directly affects the photovoltaic power generation. Typically, the area to be monitored is delineated based on images from weather radar. For details, please refer to [reference needed]. Figure 3 , Figure 3 A simplified schematic diagram of the prediction area; G i,j This represents the grid cell in the i-th row and j-th column of the current prediction region, which contains location information related to the current cloud cluster; the total number of grid cells is M×M, and its specific calculation is as follows:

[0150]

[0151] In the formula, Area represents the total area of ​​the predicted region, which is a known parameter; Indicates rounding up; M max This indicates the maximum number of grid cells constraint; set the corresponding value based on the available computing power. Size grid The actual area of ​​a single grid cell is represented by the following calculation:

[0152]

[0153] In the formula, Resolution represents the mesh size factor, which is used to control the mesh size. In this invention, it is set to 10km by default. 2 , representing the base resolution when there is no wind speed or cloud cover; The threshold for ignoring cloud cluster impact is set, with a default value of 0.0001. At that time, it was considered that the impact of the cloud clusters was negligible, and the following approach was directly adopted. As the current grid resolution factor;

[0154] λ res This represents the adaptive grid resolution adjustment factor, used to dynamically adjust the grid size based on changes in cloud velocity and area to adapt to different cloud characteristics. Its specific calculation is as follows:

[0155]

[0156] In the formula, ||u|| represents the modulus of the wind speed data, used to reflect the speed of cloud movement; u base The reference wind speed is obtained through a statistically weighted average of historical data; A cloud This indicates the cloud area in the current cloud location information; β represents the influence factor of wind speed and area change, which defaults to 0.5 and is used to balance the relative importance of wind speed and cloud coverage. This is the cloud area percentage, used to reflect the degree of cloud coverage over the predicted area.

[0157] Next, based on a preset time, a Kalman filter combined with wind speed data is used to predict the set of prediction areas after grid division, thereby obtaining the cloud cover status within the preset time; wherein, the specific iterative expression of the Kalman filter is as follows:

[0158] Let the cloud state of the grid in the i-th row and j-th column be:

[0159]

[0160] In the formula, x cloud,ij Let x be the cloud state of the i-th row and j-th column grid, used to represent the current grid's state vector; c,ij and y c,ij Represents grid G ij The location of the local cloud centroid is used to characterize the shape distribution of the cloud within the grid.

[0161] v x and v y This represents the velocity state of the cloud cluster in the i-th row and j-th column. The partial derivative, which represents the rate of change of cloud area, is used to capture cloud deformation.

[0162] Therefore, the state equation for the i-th row and j-th column grid is expressed as:

[0163]

[0164] In the formula, For grid G ij The cloud state at the predicted time k; F ij For grid G ij The state transition matrix is ​​used to determine the state change from the previous time step to the current time step, and its specific representation is as follows:

[0165]

[0166] In the formula, Δt is the time step, which defaults to 10 minutes and is used to update the cloud status; The decay term, representing the rate of change of area, is used to reflect the inertia of cloud diffusion. τ represents the cloud diffusion time constant, which is a known parameter obtained by statistical averaging of historical data.

[0167] B is the wind speed control term, used to describe the influence of observed variables such as wind speed on cloud formations. Its specific representation is as follows:

[0168]

[0169] In the formula, β x and β y The weight used to represent the influence of wind speed on cloud velocity is specifically obtained through calibration with actual data. This invention does not specifically limit its influence weight here.

[0170] This is process noise, conforming to a Gaussian distribution, used to reflect the randomness in the evolution of cloud states; u (k) As an external control input, it serves as an observed variable, i.e. The continuous velocity components of wind speed in the x and y directions are represented in this invention and are obtained by prediction using LSTM combined with historical data. The specific representation is as follows:

[0171]

[0172] In the formula, u (init) This represents the historical wind speed data from the initial measurement, used to characterize the past state of wind speed, and is obtained by measuring the echo of ultrasonic waves from weather radar; u (0) This represents the currently measured wind speed data, used to characterize the current state of the wind speed, and is obtained by measuring the echo of ultrasonic waves from a weather radar; u (k-1)This represents the predicted wind speed data at time k-1, used to characterize the future state of wind speed. It is obtained by combining LSTM with historical data for prediction. 风速 The long short-term memory recurrent neural network for predicting wind speed is trained and fitted using historical data as the training dataset. It is used to obtain the predicted value of the continuous velocity components of the wind speed in the x and y directions at time k by inputting the sequence of historical wind speed components.

[0173] A ij Indicates the relationship between neighboring grids and the current grid G. ij The influence weight matrix is ​​used to represent the degree to which the current grid's state is influenced by neighboring grids. This represents the influence matrix of neighboring grids on the current grid, specifically as follows:

[0174]

[0175] In the formula, This represents the state vector of the neighboring grid at time k+1, containing only the velocity state of the cloud; v x,(i+m,j+n) and v y,(i+m,j+n) Used to represent grid G (i+m,j+n) The corresponding cloud velocity state; R represents the influence range, which indicates the range of neighboring grids that affect the current grid, and its specific calculation is as follows:

[0176]

[0177] In the formula, base represents the baseline influence range, which is used to represent the basic influence range, and its default value is 1; This represents the rate of change of cloud area in the current grid at time k-1; γ is the area change rate sensitivity coefficient, used to represent the degree of influence of the current grid's cloud area change rate on the affected area, with a default value of 0.2.

[0178] α mn The grid influence weights, representing the horizontal and vertical offsets, are used to indicate the degree of influence of the velocity state of a cloud cluster in a grid offset by m on the horizontal axis and n on the velocity state of the cloud cluster in the current grid. The specific calculation is as follows:

[0179]

[0180] In the formula, distance mn This is a Gaussian decay weight value used to represent the degree of decay of the influence of the grid with a horizontal axis offset of m and a vertical axis offset of n on the current grid. Its specific representation is as follows:

[0181]

[0182] In the formula, [m,n] represents the grid offset, used to indicate the positional offset relative to the current grid; Δx and Δy represent the actual side length of a single grid, i.e. [m·Δx,n·Δy] is used to represent the actual position offset of the neighborhood grid [i+m,i+n] relative to the current grid; σ represents the cloud displacement vector affected by wind speed; mn The standard deviation is used to control the magnitude of the grid's influence when the horizontal axis is offset by m and the vertical axis by n. Its specific calculation is expressed as follows:

[0183]

[0184] In the formula, σ base δ is the baseline standard deviation, with a default value of 0.5km, obtained through testing based on experiments and historical data; δ is the standard deviation sensitivity coefficient, calibrated based on experiments, used to control the degree of influence of cloud area changes in a grid with a horizontal axis offset of m and a vertical axis offset of n on the current grid.

[0185] Accordingly, the neighborhood influence weight depends on the matching degree between the actual displacement of the cloud and the spatial position of the grid. When the wind speed direction is consistent with the neighborhood direction, the weight is higher. Thus, by using the information of the neighborhood grid, the Kalman filter can make a smoother and more consistent prediction of the current grid under the influence of multiple neighboring grids, thereby reducing the prediction error caused by insufficient local information.

[0186] For the grid G ​​at time k+1 ij The Kalman filter prediction of the cloud state is specifically represented as follows:

[0187]

[0188] In the formula, Represents grid G ij The predicted cloud state at time k+1 is obtained by prediction based on the posterior state estimate at time k. For grid G ij The prediction error covariance matrix is ​​used to represent the uncertainty of the predicted state at time k+1. Represents grid G ij The posterior error covariance matrix at time k, i.e., the prediction error covariance matrix updated after observation; Q ij Represents grid G ij The process noise covariance matrix is ​​used to represent the influence of external uncertainties on cloud state during the prediction process, thereby avoiding excessive reliance on historical data in the prediction process.

[0189] When the prediction obtains the grid G ​​at time k+1ij After determining the cloud state, a Kalman filter update is performed, as shown below:

[0190]

[0191] In the formula, Represents grid G ij The Kalman filter matrix is ​​used to adjust the correction amount for the predicted state; H ij Represents grid G ij The observation matrix is ​​used to map state variables to the observation space, thereby ensuring that the observation error can be calculated correctly; R represents the covariance matrix of the observation noise, which is used to represent the uncertainty of the observation data. This represents the pseudo-observation value at time k+1 obtained through LSTM prediction, used to replace the real observation data, thereby enabling Kalman filtering to perform more accurate state updates during long-term prediction. Its specific representation is as follows:

[0192]

[0193] In the formula, This represents the initial measured cloud state, and is used to characterize historical data about the cloud state. This represents the actual cloud state measured at the moment, used to characterize the true cloud state measured at the moment; LSTM 云团 To predict cloud state, a long short-term memory recurrent neural network is trained and fitted using historical data as the training dataset. This network is used to obtain the pseudo-observation value of cloud state at time k+1 by inputting the sequence of measured cloud state and predicted cloud state.

[0194] Finally, after completing the Kalman filter update within the preset time period, the specific calculation representation of the global cloud coverage state within the preset time period at time k is as follows:

[0195]

[0196] In the formula, Cover (k) Used to represent the cloud cover state at time k, that is, the cloud area in the prediction region at time k; Represents grid G ij The initial cloud area is obtained by cropping the cloud area corresponding to the current cloud location information; Represents grid G ij The Kalman filter predicts the rate of change of area at time r.

[0197] Accordingly, this invention uses LSTM wind speed prediction as a nonlinear but time-dependent input control variable to improve the long-term prediction of cloud movement trajectories by Kalman filtering under extreme weather conditions, and reduces the accumulated bias of Kalman filtering in the long-term prediction process, thereby significantly improving the accuracy of cloud coverage state prediction.

[0198] Furthermore, by employing the weighted influence term of neighboring grids, the Kalman filter's state equation is enhanced to model the changes in cloud state within a local area, avoiding the accumulation of prediction errors from a single grid from affecting the overall cloud prediction, thereby improving the accuracy of cloud cover state prediction. At the same time, by dynamically adjusting the weighted influence term of neighboring grids, the Kalman filter can adapt to cloud movement at different scales during the prediction process, thereby improving prediction adaptability.

[0199] Since Kalman filtering traditionally relies on real measurements as observations for updating, but the real measurements are limited in the long-term prediction process, this invention uses LSTM to predict cloud states and generate pseudo-observations to replace real observation data, thereby significantly enhancing the long-term prediction capability of Kalman filtering. Furthermore, since LSTM introduces nonlinear and time-dependent characteristics, it significantly improves the prediction accuracy of cloud cover states.

[0200] The cloud occlusion probability calculation unit 3 is used to perform step S3: using an occlusion probability function to calculate the cloud coverage state within a preset time period, and obtain the cloud occlusion probability within the preset time period.

[0201] Specifically, the cloud occlusion probability at time k is calculated using the occlusion probability function as follows:

[0202]

[0203] In the formula, This represents the probability of cloud cover at time k. This represents the historical occlusion probability term, used to indicate the impact of the historical occlusion probability on the current time k. The historical occlusion weight at time k is defined as follows: The lower the value, the more suitable it is for predicting rapidly changing clouds. Its specific calculation is expressed as follows:

[0204]

[0205] In the formula, β represents the rate of change of cloud cover area at time k, used to represent the cloud diffusion rate; cover This represents the historical impact adjustment coefficient, used to control the impact of the rate of change on the weights; its default value is 0.1.

[0206] This indicates the degree of coverage of the current cloud cluster over the predicted area; Let represent the cloud diffusion probability gain at time k, used to correct for the cloud diffusion effect during drastic changes. Its specific calculation is as follows:

[0207]

[0208] In the formula, γ diffusion This represents the diffusion effect adjustment coefficient, with a default value of 0.5; This indicates the degree of influence of the rate of change of saturation. The numerical range is limited by tanh to prevent the correction term from growing indefinitely. Its θ is the saturation threshold, which is used to adjust the amplitude effect when the rate of change of area exceeds the threshold. Its default value is 0.5. This represents the remaining space of the predicted area, i.e., the space that can be diffused, used to calculate the corrected occlusion probability when ensuring that there is diffuseable space.

[0209] Accordingly, this invention establishes a nonlinear probability equation driven by the rate of change of coverage area to integrate historical occlusion probability terms with the limited cloud diffusion probability gain terms, thereby achieving accurate dynamic modeling of cloud occlusion probability. At the same time, by adaptively adjusting the historical occlusion weight, the dependence on historical data is reduced in scenarios with rapid cloud changes, such as typhoons and marine climates, thereby enhancing the ability to predict cloud occlusion under short-term and drastic changes and effectively reducing error accumulation.

[0210] Furthermore, this invention also introduces a cloud diffusion probability gain to adaptively correct the occlusion probability, ensuring that the prediction can be dynamically adjusted with the cloud diffusion under rapidly spreading cloud conditions, reducing the lag in occlusion prediction. At the same time, it ensures that the cloud is close to complete coverage, and the cloud diffusion probability gain can automatically decay, avoiding the accumulation of calculation errors, thereby improving the accuracy of cloud occlusion probability calculation under severe weather conditions such as strong convection or typhoons.

[0211] The light intensity attenuation prediction unit 4 is used to perform step S4: using a light intensity attenuation model to predict the cloud coverage state and cloud occlusion probability within a preset time period, and obtain the light intensity prediction value within the preset time period.

[0212] Specifically, the calculation of the predicted light intensity value at time k obtained through the light intensity attenuation model is expressed as follows:

[0213]

[0214] In the formula, I (k) Ik represents the predicted light intensity at time k; I0 represents the current clear-sky radiation intensity, used to represent the maximum solar radiation under cloudless conditions, and is a constant obtained statistically from historical measurement data of the corresponding region. The term represents the effective optical path length, used to characterize the direct shading effect of cloud cover on the solar radiation path. area The cloud area attenuation coefficient is calculated as follows:

[0215]

[0216] In the formula, α type The cloud type coefficient is obtained through statistical calculations using historical measurement data. Its value is related to the currently predicted cloud type, and this invention does not specifically limit the value of the cloud type coefficient; h cloud The cloud thickness is indicated by measurements taken using weather radar; h ref This represents the reference height of the cloud cluster, which is a constant.

[0217] This represents the transmittance correction term, used to describe the attenuation effect of cloud thickness and occlusion probability on light intensity. Its μ represents the weighting coefficient for cloud thickness, and its specific calculation is as follows:

[0218] μ=μ0·(1-exp(-δ μ ·h cloud ))

[0219] In the formula, μ0 represents the basic optical thickness coefficient, which is obtained by inverting the weather radar reflectivity factor with an empirical formula and is a constant; δ μ The thickness sensitivity coefficient is obtained through Mie scattering calculations.

[0220] d eff This represents the equivalent path length correction factor, used to correct the geometric influence of the solar altitude angle on the cloud penetration path, to represent the equivalent path length of light rays passing through the cloud. Its specific calculation is as follows:

[0221]

[0222] In the formula, Let represent the solar altitude angle at time k, which is a known parameter.

[0223] Accordingly, the light intensity attenuation model of the present invention separates the cloud coverage area term from the shading probability term, thereby distinguishing the different effects of the geometric shading effect and the optical transmission effect of clouds on light intensity, thus improving the adaptability to light attenuation under different weather conditions. In addition, the present invention introduces an equivalent path length correction coefficient to ensure the accuracy of light intensity prediction at different time periods, thereby reducing the accumulation of errors in the calculation process and improving the accuracy of light intensity prediction.

[0224] Furthermore, the light intensity attenuation model also includes diffusion compensation and historical smoothing calculations, which are specifically represented as follows:

[0225]

[0226] In the formula, This represents the final predicted light intensity value at time k. This represents the historical compensation term, used to smooth illumination predictions and avoid short-term fluctuations affecting the overall trend. hist The historical light intensity compensation weighting coefficient is calculated as follows:

[0227]

[0228] In the formula, λ hist This represents the attenuation rate of historical data influence, used to control the degree of influence of historical data on prediction; its default value is 0.2. This represents the rate of change of the occlusion probability at time k; This represents the diffusion compensation term, used to compensate for local light leakage caused by the rapid movement of the cloud. Its specific calculation is as follows:

[0229]

[0230] In the formula, θ v This represents the threshold for the rate of change of cloud coverage area, used to measure whether the current rate of change of cloud area needs to be calculated for diffusion compensation. Its value is obtained by statistical calculation based on historical data. When the rate of change of cloud area is greater than the threshold for the rate of change of cloud coverage area, it means that the cloud is moving rapidly or the edge is becoming sparse, which may cause sunlight to pass through the gaps in the cloud layer for a short period of time, resulting in a temporary increase in the actual light intensity. Therefore, diffusion compensation is required.

[0231] Accordingly, the present invention further supplements the light intensity prediction model with a historical compensation term to reduce the impact of short-term fluctuations on light intensity prediction; and by combining a diffusion compensation term, it improves the prediction accuracy of fast-moving clouds and reduces the impact of short-term light leakage, thereby significantly improving the accuracy of light intensity prediction.

[0232] The light intensity-to-power conversion unit 5 is used to perform step S5: using a photoelectric power linear model to convert the light intensity prediction value within a preset time period to obtain the photovoltaic power generation prediction value within the preset time period.

[0233] Specifically, the photovoltaic power prediction value at time k is obtained by using the aforementioned linear photovoltaic power model for conversion, as shown below:

[0234]

[0235] In the formula, η represents the predicted photovoltaic power generation at time k; pvThe overall efficiency coefficient of a photovoltaic module is used to represent the ratio of electrical energy that can be converted per unit of light intensity, and is a known parameter; A pv The total effective photovoltaic units of the photovoltaic power station correspond to the prediction area and are known parameters; ∈ loss The photovoltaic power generation loss term is used to correct for power generation losses caused by environmental factors. Generally, a linear regression method is used to fit historical data of photovoltaic power plants to construct the photovoltaic power generation loss term. However, since the power generation loss term of different photovoltaic power plants may be affected by factors such as environment, hardware aging or climate, this invention does not specifically limit the specific method of obtaining the photovoltaic power generation loss term.

[0236] Furthermore, the light intensity attenuation model also includes historical smoothing calculation, which is specifically represented as follows:

[0237]

[0238] In the formula, This represents the final predicted photovoltaic power generation at time k. This represents the historical smoothing compensation term, used to smooth the predicted photovoltaic power generation; w pv The dynamic smoothing weight for power generation is used to determine the degree of influence of historical photovoltaic power generation forecasts on current photovoltaic power generation forecasts. Its specific calculation is expressed as follows:

[0239]

[0240] In the formula, λ pv This is the historical smoothing adjustment coefficient for photovoltaic power generation, used to control the impact of drastic changes in a short period of time. Its default value is 0.4. This represents the rate of change in power generation.

[0241] Accordingly, this invention incorporates historical smoothing terms to ensure that the predicted photovoltaic power generation exhibits a certain transitional effect, i.e., a gradual decrease or a gradual increase. This is because photovoltaic materials themselves have a certain lag effect on changes in sunlight. Therefore, by using historical data to smooth the predicted photovoltaic power generation, this invention can ensure that the predicted value is more consistent with the gradual change trend of the actual power generation, thereby avoiding scheduling errors caused by prediction errors.

[0242] The power supply and demand calculation unit 6 is used to perform step S6: calculate the difference between the predicted power load curve within a preset time period and the corresponding predicted photovoltaic power generation value to obtain the predicted power supply and demand within the preset time period.

[0243] Specifically, the method for obtaining the electricity load prediction curve within the preset time period includes the following steps:

[0244] S61. Sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence;

[0245] Specifically, the i-th normalized electricity load data in the load data sequence The specific calculation is expressed as follows:

[0246]

[0247] In the formula, L i This represents the i-th historical electricity load data within the window; The set of historical electricity load data is represented as follows:

[0248]

[0249] In the formula, i base The index represents the current time point, i.e., the most recently acquired electricity load data; range represents the size of the sliding window; where the step size of the electricity load data is Δt, which is consistent with the time step size of the aforementioned photovoltaic power generation.

[0250] S62A. Perform statistical calculations on the electricity load data sequence to obtain the statistical feature vector of the electricity load data;

[0251] Specifically, the statistical feature vector of the electricity load data is represented as follows:

[0252]

[0253] In the formula, μ win ρ represents the average value of electricity load data, used to reflect the overall level of electricity load; peak σ represents the peak-to-valley difference rate, used to measure the difference between peak and valley electricity loads; win The standard deviation of electricity load data is used to reflect the volatility of electricity load.

[0254] S62B: One-dimensional Haar wavelet is used to perform wavelet decomposition on the power load data sequence to obtain the high and low frequency feature vectors of the power load data.

[0255] Specifically, the high and low frequency feature vectors f of the electricity load data wave The specific representation is as follows:

[0256]

[0257] In the formula, μ(wave) high μ(wave) is the mean of the high-frequency components of the electricity load data, used to quantify sudden fluctuations and reflect sudden changes in electricity load; low) represents the mean of the low-frequency component of the electricity load data, used to quantify long-term trends and represent periodic electricity load characteristics.

[0258] S63. The statistical feature vector and the high- and low-frequency feature vectors of the electricity load data are spliced ​​and fused to obtain the fused feature vector;

[0259] Specifically, the fused feature vector f fusion The specific representation is as follows:

[0260]

[0261] S64. Use the electricity load regression model to predict the fused feature vector and obtain the electricity load prediction curve within a preset time period.

[0262] Specifically, the electricity load regression model is represented as follows:

[0263]

[0264] In the formula, The electricity load forecast curve for a preset time period is shown below:

[0265]

[0266] In the formula, Let represent the predicted electricity load at time k; ...

[0267]

[0268] In the formula, b k β represents the bias term at time k, used to adjust the overall bias of the electricity load forecast at time k, thereby correcting systematic errors in historical data; β represents the weight matrix of the electricity load regression model, used to control the influence of different features on the electricity load forecast, obtained by training the electricity load regression model with historical data, and its specific representation is as follows:

[0269]

[0270] In the formula, β k,j This represents the weight value in the j-th dimension at time k, where the feature dimension j defaults to 5.

[0271] Next, after obtaining the electricity load prediction curve within a preset time period, the specific representation of the difference calculation at time k is as follows:

[0272]

[0273] In the formula, η represents the predicted supply and demand power at time k, used to represent the relationship between the predicted demand and the predicted photovoltaic power generation; gird This represents the power transmission loss coefficient, used to indicate the conversion loss when photovoltaic power is transmitted to a microgrid, local energy system, or power grid. It is an empirical coefficient with a default value of 0.95; δ safe This is a safety margin, a buffer value used to ensure system safety, and its specific calculation is expressed as follows:

[0274] δ safe =λ safe ·μ(wave high )+γ safe ·σ(wave high )

[0275] In the formula, λ safe γ represents the fluctuation intensity coefficient, used to reserve a certain average high-frequency fluctuation as a buffer coefficient. It is obtained statistically based on historical accident risks and can be further calculated using time-step exponential decay. This invention does not specifically limit the selection of its fluctuation intensity coefficient; safe σ represents the uncertainty coefficient, used to assess the uncertainty of electricity consumption; high () represents the standard deviation of the high-frequency components of the electricity load data, used for safety redundancy in case of abnormal load changes, to prevent supply and demand imbalance caused by sudden changes in the system.

[0276] Accordingly, this invention introduces wavelet features on the basis of global statistical features to enhance the ability to identify sudden drops or changes in power consumption, thereby improving the accuracy of predicting sudden changes in power load. At the same time, it combines safety margin compensation in power supply and demand prediction to ensure that the system still has sufficient power buffer in the event of drastic short-term power load fluctuations, avoiding power supply fluctuations caused by excessive differences between supply and demand, and reducing the risk of power supply instability.

[0277] The predictive power supply scheduling unit 7 is used to execute step S7: scheduling and allocating the energy storage system according to the predicted power supply and demand within a preset time period, thereby completing the predictive power supply scheduling for electricity consumption.

[0278] Specifically, the scheduling allocation is represented as follows:

[0279]

[0280] In the formula, ER(k) represents the surplus ratio at time k, which is used to characterize the current sufficiency of photovoltaic power generation. Its specific calculation is as follows:

[0281]

[0282] Where, -θ store The surplus threshold is set empirically. When the surplus ratio exceeds this threshold, the current photovoltaic power generation is considered sufficient, and a portion of the generated electricity is directly stored in the energy storage system. SR(k) represents the supply-demand gap ratio, and its specific calculation is as follows:

[0283]

[0284] Where, θ critical This indicates the power deficit threshold, used to determine whether the current power deficit ratio is too low. If so, the energy reserves of the energy storage system will be used first.

[0285] Accordingly, this invention constructs predictive supply and demand power curves as the basis for decision-making in preventive dispatching strategies, enabling early identification of potential power shortages or surpluses in future periods. This guides the energy storage system to conduct proactive dispatching and energy storage management, significantly improving the response efficiency and control accuracy of dispatching strategies and reducing the risk of power system fluctuations caused by supply and demand imbalances.

[0286] It is important to note that the predicted supply and demand power within the preset time period on which the scheduling allocation is based mainly selects several short periods (e.g., 1 to 2 prediction time steps) from the current moment as the effective scheduling basis to ensure that the energy storage system completes energy allocation before a load gap appears. As for the prediction results of more distant time steps, due to their higher uncertainty, they are usually not directly involved in the scheduling calculation. They can only be used as a reference for the scheduling system in some scenarios to assist in energy storage or power purchase decisions. In the case of increasingly severe weather, the prediction period can be further increased to obtain the corresponding predicted supply and demand power, so as to achieve rolling prediction.

[0287] Meanwhile, because different energy storage systems have their own scheduling strategies and control logic, and their scale, energy storage capacity, and charging and discharging efficiency vary significantly, this invention does not limit the specific scheduling details of energy storage systems. In particular, the core of this invention lies in providing a predictive power supply and demand calculation method, and providing reasonable power dispatch suggestions based on the prediction results, so that energy storage systems can make corresponding adjustments. Therefore, the specific execution method can be adapted and optimized according to the actual energy storage system's capacity, scheduling algorithm, discharge strategy, and grid feedback capability.

[0288] Compared to existing technologies, this invention, under conditions of severe climate change, uses Kalman filtering combined with a light intensity attenuation model to accurately predict future photovoltaic power generation. It also combines historical data to perform intelligent regression prediction of electricity load, thereby calculating predictive supply and demand power. Based on this predictive supply and demand power, predictive power dispatch is carried out, effectively responding to power fluctuations under sudden weather conditions and reducing power shortages or power waste, thus significantly improving the stability and security of power supply.

[0289] Based on the same inventive concept, this application also provides an electronic device, which can be a server, desktop computing device, or mobile computing device (e.g., laptop computing device, handheld computing device, tablet computer, netbook, etc.). The device includes one or more processors and a memory, wherein the processor is used to execute a program to implement the intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand forecasting according to embodiments of the present invention; the memory is used to store computer programs executable by the processor.

[0290] Based on the same inventive concept, this application also provides a computer-readable storage medium corresponding to the aforementioned embodiments of the photovoltaic power generation intelligent scheduling method based on dynamic supply and demand forecasting. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the photovoltaic power generation intelligent scheduling method based on dynamic supply and demand forecasting described in any of the above embodiments.

[0291] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0292] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A method for intelligent scheduling of photovoltaic power generation based on dynamic supply and demand forecasting, characterized in that, Includes the following steps: S1. Obtain the current cloud cluster's location information; wherein, the cloud cluster's location information includes the cloud cluster's centroid location, cloud cluster area, cloud cluster's velocity state vector, and cloud layer thickness; S2. Combine the current cloud location information with wind speed data, and use Kalman filtering to perform gridded prediction to obtain the cloud coverage status within a preset time period; the cloud coverage status is the cloud area within the prediction area. S3. Use an occlusion probability function to calculate the cloud coverage state within a preset time period and obtain the cloud occlusion probability within the preset time period. S4. Use a light intensity attenuation model to attenuate and predict the cloud coverage status and cloud occlusion probability within a preset time period, and obtain the light intensity prediction value within the preset time period. S5. A photoelectric power linear model is used to convert the predicted light intensity value within a preset time period to obtain the predicted photovoltaic power generation value within the preset time period. S6. Calculate the difference between the predicted electricity load curve and the corresponding predicted photovoltaic power generation within a preset time period to obtain the predicted supply and demand power within the preset time period. S7. Based on the predicted power supply and demand within a preset time period, the energy storage system is scheduled and allocated to complete the predictive power supply scheduling for power consumption.

2. The intelligent dispatching method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 1, characterized in that, Step S2 includes the following sub-steps: S21. Based on the cloud area and wind speed data in the current cloud location information, the prediction area is gridded to obtain a set of prediction areas after gridding. The specific representation of the predicted region set after gridding is as follows: In the formula, This refers to the set of predicted regions after gridding, where each predicted region represents the area of ​​light intensity variation caused by the projection of cloud masses moving across the sky. Indicates the first [number] in the current prediction region. Line number The grid consists of columns, including the current location information of the cloud cluster; and the total number of grids is [number missing]. The specific calculation is expressed as follows: In the formula, This represents the total area of ​​the predicted region; Indicates rounding up; This indicates a maximum number of grid cells constraint; The actual area of ​​a single grid cell is represented by the following calculation: In the formula, Indicates the mesh size factor; The threshold for ignoring the impact of cloud clusters; The adaptive grid resolution adjustment factor is calculated as follows: In the formula, This represents the modulus of the wind speed data; The baseline wind speed is obtained through a statistically weighted average of historical data. This indicates the cloud area in the current cloud location information; Factors influencing wind speed and area changes; This refers to the percentage of cloud area. S22. Based on the preset time, a Kalman filter is used in conjunction with wind speed data to predict the set of prediction areas after grid division, so as to obtain the cloud coverage status within the preset time.

3. The intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 2, characterized in that, The specific iterative expression of the Kalman filter is as follows: No. Line number The cloud state of the column grid is as follows: In the formula, For the first Line number The cloud formation status of the grid; and Represents a grid The location of the centroid of a local cloud cluster; and Indicates the first Line number The velocity status of the cloud clusters in the grid; This represents the rate of change of the cloud area; For the Line number The state equations for the grid are expressed as follows: In the formula, For grid At the predicted time The state of the cloud formation; For grid The state transition matrix is ​​specifically represented as follows: In the formula, For time step; The attenuation term represents the rate of change of area. Represented by the cloud diffusion time constant; For wind speed control, its specific representation is as follows: In the formula, and Used to represent the weight of the influence of wind speed on cloud velocity; This is process noise; External control input, i.e., observed variables; Indicates wind speed at and The continuous velocity components in the direction are specifically obtained by prediction using LSTM combined with historical data, and are represented as follows: In the formula, This represents the historical wind speed data from the initial measurement; This indicates the currently measured wind speed data; express Predicted wind speed data at any given time; To obtain the long short-term memory recurrent neural network for predicting wind speed, historical data was used as the training dataset to train and fit the network. Indicates the neighboring grid to the current grid The influence weight matrix; The matrix representing the influence of neighboring grids on the current grid is as follows: In the formula, express The state vector of the neighboring grid at time t; and Used to represent a grid The corresponding cloud velocity status; The scope of influence is indicated, and its specific calculation is as follows: In the formula, Indicates the scope of influence of the benchmark; express The rate of change of cloud area in the current grid at any given time; This is the sensitivity coefficient for the rate of change of area; The weights representing the influence of grid offsets in the horizontal and vertical directions are calculated as follows: In the formula, The Gaussian decay weight value is represented as follows: In the formula, This is the grid offset; and This represents the actual side length of a single grid cell, i.e. ; Used to represent a domain grid Offset relative to the actual position of the current grid; This represents the cloud displacement vector affected by wind speed. The standard deviation is calculated as follows: In the formula, The baseline standard deviation; The standard deviation sensitivity coefficient; For the Time Grid The Kalman filter prediction of the cloud state is specifically represented as follows: In the formula, Represents a grid In Predicted cloud conditions at any given time; For grid The prediction error covariance matrix; Represents a grid exist The posterior error covariance matrix at time t; Represents a grid The process noise covariance matrix; When the prediction obtains the first Time Grid After determining the cloud state, a Kalman filter update is performed, as shown below: In the formula, Represents a grid The Kalman filter matrix; Represents a grid The observation matrix; The covariance matrix representing the observation noise; Indicates the first prediction obtained through LSTM. The pseudo-observation value at time 1 is used to replace the real observation data, and its specific representation is as follows: In the formula, This indicates the initial cloud state. This indicates the current measured actual cloud state; A long short-term memory recurrent neural network for predicting cloud states; Among them, when the Kalman filter update is completed within the preset time, then for the th The specific calculation representation of the cloud cover status within a preset global time period at any given moment is as follows: In the formula, Used to indicate the first The cloud cover status corresponding to the current moment, i.e., the first moment. The area of ​​cloud clusters within the forecast region at all times; Represents a grid The initial cloud area; Represents a grid Kalman filtering in the first The rate of change of area predicted at any given time.

4. The intelligent dispatching method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 3, characterized in that, The occlusion probability function is used to calculate the first... The specific representation of the cloud occlusion probability at any given time is as follows: In the formula, Indicates the first The probability of cloud cover at any given moment; This represents the historical occlusion probability term; For the first The historical occlusion weight at any given time is calculated as follows: In the formula, Indicates the first The rate of change of the area covered by the cloud at any given time; Indicates the historical impact adjustment coefficient; This indicates the degree of coverage of the current cloud cluster over the predicted area; Indicates the first The cloud diffusion probability gain at time t is calculated as follows: In the formula, Indicates the diffusion effect adjustment coefficient; This indicates the degree of influence of the rate of change of saturation. This is the saturation threshold. This represents the remaining space of the predicted area, i.e., the space for expansion.

5. The intelligent dispatching method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 4, characterized in that, For the first obtained by the light intensity attenuation model The specific calculation method for the predicted light intensity at a given time is as follows: In the formula, Indicates the final number Predicted light intensity at any given time; This indicates historical compensation items, which The historical light intensity compensation weighting coefficient is calculated as follows: In the formula, This indicates the impact of historical data on the decay rate; Indicates the first The rate of change of occlusion probability at any given time; The diffusion compensation term is represented, and its specific calculation is as follows: In the formula, The threshold representing the rate of change in the coverage area of ​​a cloud cluster; The first without historical compensation The predicted light intensity at any given time is represented as follows: In the formula, This indicates the current clear-sky radiation intensity; This represents the effective optical path term, which The cloud area attenuation coefficient is calculated as follows: In the formula, Indicates the cloud type coefficient; Indicates cloud thickness; Indicates the reference height of the cloud cluster; This represents the transmittance correction term, which... The weighting coefficient representing cloud thickness is calculated as follows: In the formula, Indicates the basic optical thickness coefficient; The thickness sensitivity coefficient is obtained through Mie scattering calculations. The equivalent path length correction factor is calculated as follows: In the formula, Indicates the first The altitude angle of the sun at any given moment.

6. The intelligent scheduling method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 5, characterized in that, The photoelectric power linear model is used for conversion to obtain the first... The specific representation of the photovoltaic power generation forecast at any given time is as follows: In the formula, Indicates the final number Predicted photovoltaic power generation at any given time; The first line represents the smoothed line without history. The predicted photovoltaic power generation at any given time is expressed as follows: In the formula, This represents the overall efficiency coefficient of a photovoltaic module; The total effective photovoltaic units of a photovoltaic power station; This represents the power loss term in photovoltaic power generation; Indicates the historical smoothing compensation term; The dynamic smoothing weight for power generation is calculated as follows: In the formula, This is the historical smoothing adjustment coefficient for photovoltaic power generation. This represents the rate of change in power generation.

7. The intelligent dispatching method for photovoltaic power generation based on dynamic supply and demand forecasting according to claim 6, characterized in that, The specific method for obtaining the electricity load prediction curve within the preset time period includes the following steps: S61. Sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence; S62A. Perform statistical calculations on the electricity load data sequence to obtain the statistical feature vector of the electricity load data; S62B: One-dimensional Haar wavelet is used to perform wavelet decomposition on the power load data sequence to obtain the high and low frequency feature vectors of the power load data. S63. The statistical feature vector and the high- and low-frequency feature vectors of the electricity load data are spliced ​​and fused to obtain the fused feature vector; S64. Use the electricity load regression model to predict the fused feature vector and obtain the electricity load prediction curve within a preset time period. Among them, after obtaining the electricity load prediction curve within a preset time period, for the first... The specific representation of the difference calculation at time point is as follows: In the formula, Indicates the first Predicted supply and demand power at any given time; The first element in the electricity load forecast curve represents the... Forecasted electricity load at any given time; Indicates the power transmission loss coefficient; To provide a safety margin, the specific calculation is expressed as follows: In the formula, Indicates the fluctuation intensity coefficient; Indicates the uncertainty coefficient; This represents the mean of the high-frequency components in the high- and low-frequency feature vectors of the electricity load data. It represents the standard deviation of the high-frequency component in the high- and low-frequency feature vectors of electricity load data.

8. A photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting, characterized in that, It includes a cloud cluster location information acquisition unit, a cloud cluster location prediction unit, a cloud cluster occlusion probability calculation unit, a light intensity attenuation prediction unit, a light intensity-to-power conversion unit, a power supply and demand calculation unit, and a predictive power supply scheduling unit; The cloud cluster location information acquisition unit is used to acquire the current cloud cluster location information; wherein, the cloud cluster location information includes the cloud cluster centroid location, cloud cluster area, cloud cluster velocity state vector, and cloud layer thickness; The cloud cluster location prediction unit is used to combine the current cloud cluster location information with wind speed data, and use Kalman filtering to perform gridded prediction to obtain the cloud cluster coverage status within a preset time period; the cloud cluster coverage status is the cloud cluster area within the prediction area. The cloud occlusion probability calculation unit is used to calculate the cloud coverage state within a preset time using an occlusion probability function to obtain the cloud occlusion probability within the preset time. The light intensity attenuation prediction unit is used to use a light intensity attenuation model to predict the cloud coverage state and cloud occlusion probability within a preset time, and obtain the light intensity prediction value within the preset time. The light intensity-to-power conversion unit is used to convert the light intensity prediction value within a preset time using a photoelectric power linear model to obtain the photovoltaic power generation prediction value within the preset time. The power supply and demand calculation unit is used to calculate the difference between the electricity load prediction curve and the corresponding photovoltaic power generation prediction value within a preset time period to obtain the predicted power supply and demand within the preset time period. The predictive power supply scheduling unit is used to schedule and allocate the energy storage system according to the predicted power supply and demand within a preset time period, thereby completing the predictive power supply scheduling for electricity consumption.

9. The intelligent photovoltaic power generation dispatching device based on dynamic supply and demand forecasting according to claim 8, characterized in that, The cloud location prediction unit is also used to perform the following sub-steps: S21. Based on the cloud area and wind speed data in the current cloud location information, the prediction area is gridded to obtain a set of prediction areas after gridding. The specific representation of the predicted region set after gridding is as follows: In the formula, This refers to the set of predicted regions after gridding, where each predicted region represents the area of ​​light intensity variation caused by the projection of cloud masses moving across the sky. Indicates the first [number] in the current prediction region. Line number The grid consists of columns, including the current location information of the cloud cluster; and the total number of grids is [number missing]. The specific calculation is expressed as follows: In the formula, This represents the total area of ​​the predicted region; Indicates rounding up; This indicates a maximum number of grid cells constraint; The actual area of ​​a single grid cell is represented by the following calculation: In the formula, Indicates the mesh size factor; The threshold for ignoring the impact of cloud clusters; The adaptive grid resolution adjustment factor is calculated as follows: In the formula, This represents the modulus of the wind speed data; The baseline wind speed is obtained through a statistically weighted average of historical data. This indicates the cloud area in the current cloud location information; Factors influencing wind speed and area changes; This refers to the percentage of cloud area. S22. Based on the preset time, use a Kalman filter combined with wind speed data to predict the set of prediction areas after dividing the grid, and obtain the cloud coverage status within the preset time. The specific iterative expression of the Kalman filter is as follows: No. Line number The cloud state of the column grid is as follows: In the formula, For the first Line number The cloud formation status of the grid; and Represents a grid The location of the centroid of a local cloud cluster; and Indicates the first Line number The velocity status of the cloud clusters in the grid; This represents the rate of change of the cloud area; For the Line number The state equations for the grid are expressed as follows: In the formula, For grid At the predicted time The state of the cloud formation; For grid The state transition matrix is ​​specifically represented as follows: In the formula, For time step; The attenuation term represents the rate of change of area. Represented by the cloud diffusion time constant; For wind speed control, its specific representation is as follows: In the formula, and Used to represent the weight of the influence of wind speed on cloud velocity; This is process noise; External control input, i.e., observed variables; Indicates wind speed at and The continuous velocity components in the direction are obtained by prediction using LSTM combined with historical data, and are specifically represented as follows: In the formula, This represents the historical wind speed data from the initial measurement; This indicates the currently measured wind speed data; express Real-time predicted wind speed data; To obtain the long short-term memory recurrent neural network for predicting wind speed, historical data was used as the training dataset to train and fit the network. Indicates the neighboring grid to the current grid The influence weight matrix; The matrix representing the influence of neighboring grids on the current grid is as follows: In the formula, express The state vector of the neighboring grid at time t; and Used to represent a grid The corresponding cloud velocity status; The scope of influence is indicated, and its specific calculation is as follows: In the formula, Indicates the scope of influence of the benchmark; express The rate of change of cloud area in the current grid at any given time; This is the sensitivity coefficient for the rate of change of area; The grid influence weights representing horizontal and vertical offsets are calculated as follows: In the formula, The Gaussian decay weight value is represented as follows: In the formula, This is the grid offset; and This represents the actual side length of a single grid cell, i.e. ; Used to represent a domain grid Offset relative to the actual position of the current grid; This represents the cloud displacement vector affected by wind speed. The standard deviation is calculated as follows: In the formula, The baseline standard deviation; The standard deviation sensitivity coefficient; For the Time Grid The Kalman filter prediction of the cloud state is specifically represented as follows: In the formula, Represents a grid In Predicted cloud conditions at any given time; For grid The prediction error covariance matrix; Represents a grid exist The posterior error covariance matrix at time t; Represents a grid The process noise covariance matrix; When the prediction obtains the first Time Grid After determining the cloud state, a Kalman filter update is performed, as shown below: In the formula, Represents a grid The Kalman filter matrix; Represents a grid The observation matrix; The covariance matrix representing the observation noise; Indicates the first prediction obtained through LSTM. The pseudo-observation value at time 1 is used to replace the real observation data, and its specific representation is as follows: In the formula, This indicates the initial cloud state. This indicates the current measured actual cloud state; A long short-term memory recurrent neural network for predicting cloud states; Among them, when the Kalman filter update is completed within the preset time, then for the th The specific calculation representation of the cloud cover status within a preset global time period at any given moment is as follows: In the formula, Used to indicate the first The cloud cover status corresponding to the current moment, i.e., the first moment. The area of ​​the cloud cluster within the forecast region at any given time; Represents a grid The initial cloud area; Represents a grid Kalman filtering in the first The rate of change of area predicted at any given time; Wherein, the occlusion probability function is calculated for the first... The specific representation of the cloud occlusion probability at any given time is as follows: In the formula, Indicates the first The probability of cloud cover at any given moment; This represents the historical occlusion probability term; For the first The historical occlusion weight at any given time is calculated as follows: In the formula, Indicates the first The rate of change of the area covered by the cloud at any given time; Indicates the historical impact adjustment coefficient; This indicates the degree of coverage of the current cloud cluster over the predicted area; Indicates the first The cloud diffusion probability gain at time t is calculated as follows: In the formula, Indicates the diffusion effect adjustment coefficient; This indicates the degree of influence of the rate of change of saturation. This is the saturation threshold. This represents the remaining space, or space, of the predicted area; Among them, the first obtained by the light intensity attenuation model The specific calculation method for the predicted light intensity at a given time is as follows: In the formula, Indicates the final number Predicted light intensity at any given time; This indicates historical compensation items, which The historical light intensity compensation weighting coefficient is calculated as follows: In the formula, This indicates the impact of historical data on the decay rate; Indicates the first The rate of change of occlusion probability at any given time; The diffusion compensation term is represented, and its specific calculation is as follows: In the formula, The threshold representing the rate of change in the coverage area of ​​a cloud cluster; The first without historical compensation The predicted light intensity at any given time is represented as follows: In the formula, This indicates the current clear-sky radiation intensity; This represents the effective optical path term, which The cloud area attenuation coefficient is calculated as follows: In the formula, Indicates the cloud type coefficient; Indicates cloud thickness; Indicates the reference height of the cloud cluster; This represents the transmittance correction term, which... The weighting coefficient representing cloud thickness is calculated as follows: In the formula, Indicates the basic optical thickness coefficient; The thickness sensitivity coefficient is obtained through Mie scattering calculations. The equivalent path length correction factor is calculated as follows: In the formula, Indicates the first The altitude angle of the sun at any given moment; The photoelectric power linear model is used for conversion to obtain the first... The specific representation of the photovoltaic power generation forecast at any given time is as follows: In the formula, Indicates the final number Predicted photovoltaic power generation at any given time; The first line represents the smoothed line without history. The predicted photovoltaic power generation at any given time is expressed as follows: In the formula, This represents the overall efficiency coefficient of a photovoltaic module; The total effective photovoltaic units of a photovoltaic power station; This represents the power loss term in photovoltaic power generation; Indicates the historical smoothing compensation term; The dynamic smoothing weight for power generation is calculated as follows: In the formula, This is the historical smoothing adjustment coefficient for photovoltaic power generation. This represents the rate of change in power generation.

10. The photovoltaic power generation intelligent dispatching device based on dynamic supply and demand forecasting according to claim 8, characterized in that, The specific method for obtaining the electricity load prediction curve within the preset time period includes the following steps: S61. Sampling and normalizing historical electricity load data through a sliding window to obtain an electricity load data sequence; S62A. Perform statistical calculations on the electricity load data sequence to obtain the statistical feature vector of the electricity load data; S62B: One-dimensional Haar wavelet is used to perform wavelet decomposition on the power load data sequence to obtain the high and low frequency feature vectors of the power load data. S63. The statistical feature vector and high- and low-frequency feature vectors of the electricity load data are spliced ​​and fused to obtain a fused feature vector; S64. Use the electricity load regression model to predict the fused feature vector and obtain the electricity load prediction curve within a preset time period. Among them, after obtaining the electricity load prediction curve within a preset time period, for the first... The specific representation of the difference calculation at time point is as follows: In the formula, Indicates the first Predicted supply and demand power at any given time; The first element in the electricity load forecast curve represents the... Forecasted electricity load at any given time; Indicates the power transmission loss coefficient; To provide a safety margin, the specific calculation is expressed as follows: In the formula, Indicates the fluctuation intensity coefficient; Indicates the uncertainty coefficient; This represents the mean of the high-frequency components in the high- and low-frequency feature vectors of the electricity load data. It represents the standard deviation of the high-frequency component in the high- and low-frequency feature vectors of electricity load data.

Images