Photovoltaic power generation management method and system based on multi-source data

Abstract

The invention discloses a photovoltaic power generation management method and system based on multi-source data, and the method comprises the steps: obtaining the operation data and state data of a photovoltaic power generation unit, carrying out the data classification and abnormality screening, obtaining the abnormal data of photovoltaic power generation, determining a photovoltaic power generation unit equipment parameter management scheme according to the abnormal data of photovoltaic power generation, and carrying out the management of the photovoltaic power generation unit equipment parameter. The method comprises the steps of constructing a photovoltaic power generation prediction model, inputting operation data and state data of a photovoltaic power generation set to be predicted into the photovoltaic power generation prediction model to obtain predicted photovoltaic power generation capacity, constructing a power storage and distribution target function, and determining a photovoltaic power generation set power storage and distribution scheme according to the predicted photovoltaic power generation capacity and the power storage and distribution target function. And performing photovoltaic power generation management according to the equipment parameter management scheme and the power storage and distribution scheme. The method not only can improve the efficiency and accuracy of photovoltaic power generation management, but also has good interpretability, and can be directly applied to a photovoltaic power generation management system based on multi-source data.

Description

Technical Field

[0001] The present invention relates to the field of power generation management, and in particular to a photovoltaic power generation management method and system based on multi-source data. Background Art

[0002] In the context of global energy structure transformation and sustainable development, photovoltaic power generation has attracted widespread attention due to its clean and renewable technical advantages. This technology can reduce the consumption of fossil energy, reduce carbon emissions, and enhance society's ability to cope with climate change and energy crises. Since the operation process of photovoltaic power generation systems involves many devices and complex operating parameters, how to effectively manage and optimize the photovoltaic power generation process to improve power generation efficiency, reduce operation and maintenance costs, and ensure the stable operation of the system has become a key issue in research in this field.

[0003] At the same time, traditional photovoltaic management technology still has many shortcomings: first, the existing technology mostly relies on historical power generation data or single meteorological information, lacks the deep integration of real-time equipment status and multi-source environmental data, resulting in abnormal detection lag and insufficient prediction accuracy; second, traditional storage and distribution solutions are mostly based on fixed rules or static models, which are difficult to adapt to the temporal and spatial differences between photovoltaic output volatility and user needs, and are prone to unreasonable energy storage configuration, increased line loss and other problems; in addition, the existing prediction model has limited ability to extract multimodal data features, and the optimization algorithm is difficult to handle high-dimensional non-convex objective functions, resulting in low decision-making efficiency. In response to the above problems, the present invention realizes accurate monitoring of equipment status and parameter adjustment through multi-source heterogeneous data collaborative analysis technology, adopts a multi-modal prediction-optimization integrated architecture to improve the prediction accuracy of power generation, realize dynamic optimization of storage and distribution solutions, and constructs an objective function for full life cycle cost optimization, and proposes a photovoltaic power generation management method and system based on multi-source data fusion and dynamic collaborative optimization to overcome the shortcomings of traditional photovoltaic power generation management methods, which is conducive to the efficient and stable operation of photovoltaic power generation systems, and has positive significance for promoting technological innovation and sustainable development in the energy field. Summary of the invention

[0004] The purpose of the present invention is to provide a photovoltaic power generation management method and system based on multi-source data.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention comprises the following steps:

[0007] Acquire operation data and status data of the photovoltaic power generation group, and pre-process the operation data and the status data; the status data includes environmental status data and equipment status data;

[0008] Performing data classification and abnormal screening on the operation data and the equipment status data to obtain abnormal photovoltaic power generation data, determining a photovoltaic power generation group equipment parameter management plan according to the abnormal photovoltaic power generation data, and adjusting the photovoltaic power generation group equipment parameters;

[0009] Acquire updated data of the photovoltaic power generation group, construct a photovoltaic power generation prediction model according to the updated data and the environmental status data, and input the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain predicted photovoltaic power generation;

[0010] Construct a power storage and distribution objective function, determine a power storage and distribution plan for the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function, and manage photovoltaic power generation according to the equipment parameter management plan and the power storage and distribution plan.

[0011] Furthermore, the method for obtaining the photovoltaic power generation abnormal data includes:

[0012] Obtaining operation data and status data of a photovoltaic power generation group; the photovoltaic power generation group includes photovoltaic modules, inverters, energy storage equipment and power distribution equipment; the status data includes environmental status data and equipment status data; the environmental status data is the current meteorological information and meteorological environment information forecast provided by the Meteorological Bureau;

[0013] A hierarchical clustering method is used to classify the operation data and equipment status data of the photovoltaic power generation group to obtain operation indicators and equipment status indicators; the operation indicators include photovoltaic component operation parameters, inverter operation parameters, energy storage device operation parameters and distribution device operation parameters; the equipment status indicators include photovoltaic component status parameters, inverter status parameters, energy storage device status parameters and distribution device status parameters;

[0014] Calculating the equipment energy efficiency index of the photovoltaic power generation group according to the operation index and the equipment status index; the equipment energy efficiency index includes photovoltaic module energy efficiency parameters, inverter energy efficiency parameters, energy storage equipment energy efficiency parameters and power distribution energy efficiency parameters;

[0015] According to the hierarchical clustering results, multiple hyperspheres are constructed to cover different distribution areas of normal data corresponding to operation indicators, equipment status indicators and equipment energy efficiency indicators, and abnormal screening is performed to obtain abnormal data of photovoltaic power generation.

[0016] Furthermore, the method for determining the photovoltaic power generation group equipment parameter management scheme includes:

[0017] When the PV module status parameters are abnormal, the corresponding PV module status parameters will be adjusted according to the current meteorological information. When the PV module operation parameters and PV module energy efficiency parameters are abnormal, the PV module status parameters will be adjusted, and the shadow shielding, wiring conditions and PV panel working hours will be checked, and surface dust cleaning and heat dissipation operations will be performed;

[0018] When the inverter operating parameters and inverter status parameters are abnormal, optimize the inverter control strategy and heat dissipation strategy; when the inverter energy efficiency parameters are abnormal, repair or replace the inverter;

[0019] When the state parameters of the energy storage equipment are abnormal, adjust the corresponding state parameters of the energy storage equipment according to the operating parameters of the power distribution equipment; when the operating parameters of the energy storage equipment are abnormal, optimize the state parameters of the energy storage equipment; when the operating parameters of the energy storage equipment and the energy efficiency parameters of the energy storage equipment are abnormal, check or replace old batteries;

[0020] When the distribution equipment operates abnormally, check the faults of photovoltaic modules and inverters. When the distribution equipment operating parameters, distribution equipment status parameters and distribution equipment energy efficiency parameters are abnormal, optimize the storage and distribution plan.

[0021] Furthermore, the method for obtaining the predicted photovoltaic power generation includes:

[0022] The photovoltaic power generation group update data and the current meteorological information are combined into a comprehensive data set, and the comprehensive data set is divided into a training set and a test set according to an 8:2 ratio; the photovoltaic power generation group update data includes operation update data and equipment status update data;

[0023] Constructing a photovoltaic power generation prediction model, wherein the photovoltaic power generation prediction model includes a data preprocessing layer, a multimodal prediction layer, a strategy layer, and an output layer;

[0024] The data preprocessing layer performs spatiotemporal alignment of the input data according to the time information and geographic information, and performs standardization and feature construction on the spatiotemporal aligned data to output the corresponding feature vector;

[0025] The multimodal prediction layer predicts the corresponding photovoltaic power generation according to the equipment status and meteorological information of the photovoltaic power generation group, including a random forest base model, a long short-term memory network base model and a support vector machine base model; the random forest base model fits the nonlinear relationship between the operation data, the equipment status data and the meteorological information, and predicts the photovoltaic power generation, and the mean square error loss function is used to evaluate the model performance; the long short-term memory network base model processes the input data of time continuity, captures the dependency between the data, predicts the change trend of the photovoltaic power generation in the time series, uses the mean square error loss function to calculate the square difference between the predicted value and the actual value, and uses the Adam optimizer to adjust the learning rate; the support vector machine base model maps the equipment status data and the meteorological data to the high-dimensional space through the kernel function, finds the optimal hyperplane to predict the photovoltaic power generation, uses the ε-insensitive loss function to adjust the loss according to the error between the predicted value and the actual value, and optimizes the model parameters by solving the convex optimization problem;

[0026] The strategy layer uses stacking fusion to fuse the photovoltaic power generation prediction results of the three base models, outputs the predicted photovoltaic power generation through the output layer, and uses the test set data to verify the photovoltaic power generation prediction model;

[0027] The operation data and status data of the photovoltaic power generation group to be predicted are input into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation.

[0028] Furthermore, the method for constructing the power storage and distribution objective function includes:

[0029] The storage and distribution objective function is constructed based on the user's power supply satisfaction, the overall change rate of the distribution plan, the distribution cost and the energy storage cost. The expression is:

[0030]

[0031] Among them, F aim is the power storage and distribution objective function, w 1 is the power supply satisfaction weight, n is the number of power supply areas, w 1i is the weight coefficient of region i, is the electricity demand of region i, is the photovoltaic power supply in region i, w 2 is the change rate weight, w 21 is the vector weight, w 22 is the parameter weight, X new is the characteristic vector of the new power distribution scheme, X old is the characteristic vector of the original power distribution scheme, ‖·‖ 2 is the Euclidean distance of the feature vector, δ(x k ) is the characteristic parameter x of the power distribution scheme k The penalty value, w 3 is the distribution cost weight, m ​​is the number of distribution lines, c tran is the unit distance cost, d j is the distance of distribution line j, P j is the transmission power of distribution line j, λ is the loss cost coefficient, R j is the resistivity of distribution line j, t j is the power supply duration of distribution line j, w 4 is the energy storage cost weight, c cap is the unit capacity cost, E rate is the rated capacity of energy storage, N cycle is the nominal cycle number, η is the charge and discharge loss coefficient, is the energy storage power in the time period T The sum of the absolute values ​​of .

[0032] Furthermore, the method for determining the photovoltaic power generation group storage and distribution scheme includes:

[0033] The current user power supply demand is determined by the distribution equipment operation data, and the photovoltaic power generation group storage and distribution plan is determined by comparing the predicted photovoltaic power generation with the current user power supply demand:

[0034] When the predicted photovoltaic power generation is greater than the current user power supply demand, power distribution continues according to the original power distribution plan, and the state parameters of the power storage equipment are adjusted according to the overflow power to store power;

[0035] When the predicted photovoltaic power generation is less than the current user power supply demand, the power ratio of the predicted photovoltaic power generation to the current user power supply demand is calculated, the power distribution plan is selected according to the power ratio, and the state parameters of the power storage device are adjusted according to the power ratio for discharge;

[0036] The specific method of selecting a power distribution scheme according to the power ratio is as follows:

[0037] When the power ratio is less than or equal to 50%, all the predicted photovoltaic power generation and power storage equipment power will be allocated to the nearest power supply area, and the power demand of the remaining power supply area will be provided by the power grid;

[0038] When the power ratio is greater than 50%, the current power distribution plan is optimized, and the remaining power demand of each power supply area is provided by the power grid; the power distribution plan includes the power supply areas allocated to different photovoltaic power generation groups and power storage devices, and the allocated power obtained by the power supply areas;

[0039] The specific method for optimizing the current power distribution scheme is:

[0040] The predicted photovoltaic power generation feature vector, the current user power demand feature vector, the photovoltaic power generation group state data feature vector and the current meteorological information feature vector are combined into a first photovoltaic power generation group feature vector, and the first photovoltaic power generation group feature vector is matched with a photovoltaic power generation history database to obtain a historical power distribution plan; the photovoltaic power generation history database includes historical photovoltaic power generation group feature vectors and corresponding power distribution plans; the method for obtaining the historical power distribution plan is to calculate the comprehensive similarity between the first photovoltaic power generation group feature vector and the historical photovoltaic power generation group feature vector, and take the power distribution plan corresponding to the historical photovoltaic power generation group feature vector with the highest comprehensive similarity as the historical power distribution plan;

[0041] According to the power storage and distribution objective function, the historical power distribution plan is optimized, and the particle swarm algorithm is used to take the historical power distribution plan as the initial particle population to determine the current individual optimal position p best and the optimal position of the historical population g best , update the particle velocity and particle position:

[0042]

[0043] in is the speed of particle i in the k+1 update, is the speed of particle i in the kth update, θ is the particle speed adaptive inertia weight, q 1 ,q 2 is the learning factor, r 1 、r 2 is a random number from (0,1), is the position of particle i in the kth update, v is the regularization weight, is the position of particle i in the k+1th update, Levy is the Levy flight step length, θ max ,θ min is the maximum and minimum value of the adaptive inertia weight, K is the maximum number of updates, q 1s ,q 2s is the initial value of the learning factor, q 1e ,q 2e is the termination value of the learning factor, α, β, γ are dynamic adjustment coefficients;

[0044] Set the crossover threshold When the similarity between two bodies is greater than the crossover critical value, a crossover operation is performed:

[0045] x c =p mix x p1 +(1-p mix )x p2 +Gauss·σ

[0046]

[0047] where x c is the position of the descendant particle, p mix is the hybridization probability, x p1 、x p2 is the position of the particles of the two parents, Gauss is a Gaussian distribution random number, σ is the standard deviation of the Gaussian distribution, v c is the velocity of the descendant particle, v p1 、v p2 are the velocities of the two parent particles;

[0048] Use dynamic Cauchy mutation strategy to perform mutation operation:

[0049]

[0050] g best =p best (1+0.5Cauchy)

[0051] Where Cauchy is the Cauchy mutation strategy, p var is the adaptive mutation probability, ε is the dynamic adjustment coefficient;

[0052] Iteratively update the particle position and speed until the power storage and distribution objective function is minimized and then stop updating to obtain the optimal position g of the historical population. best The optimized power distribution plan is output, and the optimized power distribution plan and the corresponding first photovoltaic power generation group feature vector are uploaded to the photovoltaic power generation history database.

[0053] In a second aspect, a photovoltaic power generation management system based on multi-source data includes:

[0054] Data acquisition module: used to obtain the operation data and status data of the photovoltaic power generation group, and pre-process the operation data and the status data;

[0055] Equipment parameter module: used for classifying and screening the operation data and the equipment status data to obtain photovoltaic power generation abnormal data, and determining the photovoltaic power generation group equipment parameter management plan according to the photovoltaic power generation abnormal data;

[0056] Photovoltaic power generation prediction module: used to construct a photovoltaic power generation prediction model according to the update data and the environmental status data, and input the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation;

[0057] Power storage and distribution scheme module: used to construct a power storage and distribution objective function, and determine the power storage and distribution scheme of the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function;

[0058] Management module: used to store, view and manage the operation data, the status data, the abnormal photovoltaic power generation data, the equipment parameter management plan and the power storage and distribution plan, perform abnormality detection and early warning based on the abnormal photovoltaic power generation data, and perform photovoltaic power generation management based on the equipment parameter management plan and the power storage and distribution plan.

[0059] The beneficial effects of the present invention are:

[0060] The present invention is a photovoltaic power generation management method and system based on multi-source data. Compared with the prior art, the present invention has the following technical effects:

[0061] The present invention can improve the data preprocessing capability and enhance the model adaptability in photovoltaic power generation management through the steps of anomaly detection, data prediction, building objective function, data matching and search optimization, and can realize anomaly detection and precise matching of multi-source heterogeneous data, thereby improving the efficiency and accuracy of photovoltaic power generation management. The photovoltaic power generation management technology can be optimized, which can greatly save resources and improve work efficiency, realize accurate prediction and optimal control of photovoltaic power generation systems, provide reference and reference for the management and optimization of photovoltaic power generation, and have positive significance for promoting technological innovation and sustainable development in the energy field. It can adapt to the terminal management needs of photovoltaic power generation management systems and different users, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The present invention is a flowchart of the steps of a photovoltaic power generation management method based on multi-source data. DETAILED DESCRIPTION

[0063] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0064] The present invention provides a photovoltaic power generation management method and system based on multi-source data, comprising the following steps:

[0065] like Figure 1 As shown, in this embodiment, the following steps are included:

[0066] Acquire operation data and status data of the photovoltaic power generation group, and pre-process the operation data and the status data; the status data includes environmental status data and equipment status data;

[0067] Performing data classification and abnormal screening on the operation data and the equipment status data to obtain abnormal photovoltaic power generation data, determining a photovoltaic power generation group equipment parameter management plan according to the abnormal photovoltaic power generation data, and adjusting the photovoltaic power generation group equipment parameters;

[0068] Acquire updated data of the photovoltaic power generation group, construct a photovoltaic power generation prediction model according to the updated data and the environmental status data, and input the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain predicted photovoltaic power generation;

[0069] Construct a power storage and distribution objective function, determine a power storage and distribution plan for the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function, and manage photovoltaic power generation according to the equipment parameter management plan and the power storage and distribution plan.

[0070] In this embodiment, the method for obtaining the photovoltaic power generation abnormal data includes:

[0071] Obtaining operation data and status data of a photovoltaic power generation group; the photovoltaic power generation group includes photovoltaic modules, inverters, energy storage equipment and power distribution equipment; the status data includes environmental status data and equipment status data; the environmental status data is the current meteorological information and meteorological environment information forecast provided by the Meteorological Bureau;

[0072] A hierarchical clustering method is used to classify the operation data and equipment status data of the photovoltaic power generation group to obtain operation indicators and equipment status indicators; the operation indicators include photovoltaic component operation parameters, inverter operation parameters, energy storage device operation parameters and distribution device operation parameters; the equipment status indicators include photovoltaic component status parameters, inverter status parameters, energy storage device status parameters and distribution device status parameters;

[0073] Calculating the equipment energy efficiency index of the photovoltaic power generation group according to the operation index and the equipment status index; the equipment energy efficiency index includes photovoltaic module energy efficiency parameters, inverter energy efficiency parameters, energy storage equipment energy efficiency parameters and power distribution energy efficiency parameters;

[0074] According to the hierarchical clustering results, multiple hyperspheres are constructed to cover the different distribution areas of normal data corresponding to the operation indicators, equipment status indicators and equipment energy efficiency indicators, and abnormal screening is performed to obtain abnormal data of photovoltaic power generation;

[0075] In the actual evaluation, the photovoltaic power generation group in a certain place is taken as an example to obtain the operation data and status data of the photovoltaic power generation group: taking a photovoltaic module as an example, the operation parameters of the photovoltaic module include photovoltaic panel temperature of 40℃, photovoltaic panel voltage of 360V, photovoltaic panel current of 7A, the status parameters of the photovoltaic module include photovoltaic panel area of ​​1.5㎡, photovoltaic panel angle of 30°, photovoltaic panel direction of due south, photovoltaic module energy efficiency parameters include theoretical power generation of 5.6kW, actual power generation of 4.4kW, power loss of 21.4% and power change rate of 9.4% compared with the previous moment; the inverter operation parameters include input DC 390V, output AC 215V, input-output voltage ratio of 0.55 and power factor of 0.91, the inverter status parameters include control strategy (MPPT control strategy) and heat dissipation design (air cooling, speed of 1400rpm), the inverter energy efficiency parameters include AC conversion efficiency of 0.92 and power factor of 0.91; the operation parameters of the energy storage equipment include AC power conversion efficiency of 0.92 and power factor of 0.91. The operation parameters include charge and discharge state (discharging), charge and discharge cycle (400 times, 6h) and capacity (75%). The state parameters of energy storage equipment include energy storage control strategy (timing charge and discharge) and charge and discharge strategy (charging when photovoltaic power generation is greater than the power demand of users, and discharging on the contrary). The energy storage equipment energy efficiency parameters include capacity utilization rate (rated 100kWh, 75%), cycle life (rated 1500 times, remaining 73.3%), charging efficiency 90%, and discharge efficiency 85%. The operation parameters of distribution equipment include photovoltaic power generation power of 35kW and current of 90A, user power supply demand of 50kW and current of 125A. The state parameters of distribution equipment include the number and location of photovoltaic groups, user area and location, photovoltaic switch state (8 photovoltaic modules, area A demand 20kW, area B demand 15kW, area C demand 15kW, photovoltaic all closed), and the distribution energy efficiency parameters include power factor 0.94 and photovoltaic power change rate 6.1%.

[0076] Current weather information: temperature 25°C, humidity 65%, light intensity 600W / m 2 , wind speed 4m / s; meteorological environment information forecast: temperature 23-28℃, humidity 60-70%, light intensity 500-700W / m in the next 24 hours 2 , wind speed 3-5m / s;

[0077] A hypersphere was constructed based on the hierarchical clustering results, with the mean of normal data as the center. The radius of the hypersphere was determined according to the standard deviation of the data, and abnormal data was screened out: the power factor of the inverter was 0.91, which was slightly lower than the normal range; the actual power generation of the photovoltaic module was 35kW, which deviated greatly from the historical data and theoretical power generation during the same period; and the charging and discharging efficiency of the energy storage equipment suddenly dropped to 70%.

[0078] In this embodiment, the method for determining the photovoltaic power generation group equipment parameter management solution includes:

[0079] When the PV module status parameters are abnormal, the corresponding PV module status parameters will be adjusted according to the current meteorological information. When the PV module operation parameters and PV module energy efficiency parameters are abnormal, the PV module status parameters will be adjusted, and the shadow shielding, wiring conditions and PV panel working hours will be checked, and surface dust cleaning and heat dissipation operations will be performed;

[0080] When the inverter operating parameters and inverter status parameters are abnormal, optimize the inverter control strategy and heat dissipation strategy; when the inverter energy efficiency parameters are abnormal, repair or replace the inverter;

[0081] When the state parameters of the energy storage equipment are abnormal, adjust the corresponding state parameters of the energy storage equipment according to the operating parameters of the power distribution equipment; when the operating parameters of the energy storage equipment are abnormal, optimize the state parameters of the energy storage equipment; when the operating parameters of the energy storage equipment and the energy efficiency parameters of the energy storage equipment are abnormal, check or replace old batteries;

[0082] When the power distribution equipment operates abnormally, check the faults of photovoltaic modules and inverters; when the power distribution equipment operating parameters, power distribution equipment status parameters and power distribution equipment energy efficiency parameters are abnormal, optimize the power storage and distribution plan;

[0083] In the actual evaluation, the specific equipment parameter management plan based on the above abnormal data is: 1. Optimize the inverter control strategy and heat dissipation strategy; 2. Check the shadow obstruction, wiring condition and working hours of the photovoltaic panels, and perform surface cleaning and heat dissipation operations; 3. Replace aging battery packs.

[0084] In this embodiment, the method for obtaining the predicted photovoltaic power generation includes:

[0085] The photovoltaic power generation group update data and the current meteorological information are combined into a comprehensive data set, and the comprehensive data set is divided into a training set and a test set according to an 8:2 ratio; the photovoltaic power generation group update data includes operation update data and equipment status update data;

[0086] Constructing a photovoltaic power generation prediction model, wherein the photovoltaic power generation prediction model includes a data preprocessing layer, a multimodal prediction layer, a strategy layer, and an output layer;

[0087] The data preprocessing layer performs spatiotemporal alignment of the input data according to the time information and geographic information, and performs standardization and feature construction on the spatiotemporal aligned data to output the corresponding feature vector;

[0088] The multimodal prediction layer predicts the corresponding photovoltaic power generation according to the equipment status and meteorological information of the photovoltaic power generation group, including a random forest base model, a long short-term memory network base model and a support vector machine base model; the random forest base model fits the nonlinear relationship between the operation data, the equipment status data and the meteorological information, and predicts the photovoltaic power generation, and the mean square error loss function is used to evaluate the model performance; the long short-term memory network base model processes the input data of time continuity, captures the dependency between the data, predicts the change trend of the photovoltaic power generation in the time series, uses the mean square error loss function to calculate the square difference between the predicted value and the actual value, and uses the Adam optimizer to adjust the learning rate; the support vector machine base model maps the equipment status data and the meteorological data to the high-dimensional space through the kernel function, finds the optimal hyperplane to predict the photovoltaic power generation, uses the ε-insensitive loss function to adjust the loss according to the error between the predicted value and the actual value, and optimizes the model parameters by solving the convex optimization problem;

[0089] The strategy layer uses stacking fusion to fuse the photovoltaic power generation prediction results of the three base models, outputs the predicted photovoltaic power generation through the output layer, and uses the test set data to verify the photovoltaic power generation prediction model;

[0090] Inputting the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation;

[0091] In the actual evaluation, the 168 updated data per hour in the past week were divided into a training set (134 items) and a test set (34 items) according to the ratio of 8:2. The initial learning rate of the Adam optimizer of the long short-term memory network base model in the multimodal prediction layer was set to 0.001, and the loss function of the support vector machine base model was ε=0.1. The photovoltaic power generation prediction model predicted that the predicted photovoltaic power generation was 30kW.

[0092] In this embodiment, the method for constructing the power storage and distribution objective function includes:

[0093] The storage and distribution objective function is constructed based on the user's power supply satisfaction, the overall change rate of the distribution plan, the distribution cost and the energy storage cost. The expression is:

[0094]

[0095] Among them, F aim is the power storage and distribution objective function, w 1 is the power supply satisfaction weight, n is the number of power supply areas, w 1i is the weight coefficient of region i, is the electricity demand of region i, is the photovoltaic power supply in region i, w 2 is the change rate weight, w 21 is the vector weight, w22 is the parameter weight, X new is the characteristic vector of the new power distribution scheme, X old is the characteristic vector of the original power distribution scheme, ‖·‖ 2 is the Euclidean distance of the feature vector, δ(x k ) is the characteristic parameter x of the power distribution scheme k The penalty value, w 3 is the distribution cost weight, m ​​is the number of distribution lines, c tran is the unit distance cost, d j is the distance of distribution line j, P j is the transmission power of distribution line j, λ is the loss cost coefficient, R j is the resistivity of distribution line j, t j is the power supply duration of distribution line j, w 4 is the energy storage cost weight, c cap is the unit capacity cost, E rate is the rated capacity of energy storage, N cycle is the nominal cycle number, η is the charge and discharge loss coefficient, is the energy storage power in the time period T The sum of the absolute values ​​of

[0096] In the actual evaluation, the constraints of the power storage and distribution objective function are:

[0097]

[0098] Where Equation 1 is the power balance condition, For all photovoltaic power generation, is the total energy storage discharge power, For all load powers, For all energy storage charging power, is the power loss, SOC min , SOC max The minimum and maximum energy storage states are set, SOC t is the current energy storage state, P ess,min , P ess,max is the set minimum energy storage power and maximum energy storage power, P ess,t is the current energy storage power, is the transmission power of the current distribution line j, P j,max is the maximum power capacity of distribution line j, ΔV is the voltage fluctuation, V nom is the rated voltage.

[0099] In this embodiment, the method for determining the photovoltaic power generation group storage and distribution scheme includes:

[0100] The current user power supply demand is determined by the distribution equipment operation data, and the photovoltaic power generation group storage and distribution plan is determined by comparing the predicted photovoltaic power generation with the current user power supply demand:

[0101] When the predicted photovoltaic power generation is greater than the current user power supply demand, power distribution continues according to the original power distribution plan, and the state parameters of the power storage equipment are adjusted according to the overflow power to store power;

[0102] When the predicted photovoltaic power generation is less than the current user power supply demand, the power ratio of the predicted photovoltaic power generation to the current user power supply demand is calculated, the power distribution plan is selected according to the power ratio, and the state parameters of the power storage device are adjusted according to the power ratio for discharge;

[0103] The specific method of selecting a power distribution scheme according to the power ratio is as follows:

[0104] When the power ratio is less than or equal to 50%, all the predicted photovoltaic power generation and power storage equipment power will be allocated to the nearest power supply area, and the power demand of the remaining power supply area will be provided by the power grid;

[0105] When the power ratio is greater than 50%, the current power distribution plan is optimized, and the remaining power demand of each power supply area is provided by the power grid; the power distribution plan includes the power supply areas allocated to different photovoltaic power generation groups and power storage devices, and the allocated power obtained by the power supply areas;

[0106] The specific method for optimizing the current power distribution scheme is:

[0107] The predicted photovoltaic power generation feature vector, the current user power demand feature vector, the photovoltaic power generation group state data feature vector and the current meteorological information feature vector are combined into a first photovoltaic power generation group feature vector, and the first photovoltaic power generation group feature vector is matched with a photovoltaic power generation history database to obtain a historical power distribution plan; the photovoltaic power generation history database includes historical photovoltaic power generation group feature vectors and corresponding power distribution plans; the method for obtaining the historical power distribution plan is to calculate the comprehensive similarity between the first photovoltaic power generation group feature vector and the historical photovoltaic power generation group feature vector, and take the power distribution plan corresponding to the historical photovoltaic power generation group feature vector with the highest comprehensive similarity as the historical power distribution plan;

[0108] According to the power storage and distribution objective function, the historical power distribution plan is optimized, and the particle swarm algorithm is used to take the historical power distribution plan as the initial particle population to determine the current individual optimal position p best and the optimal position of the historical population g best , update the particle velocity and particle position:

[0109]

[0110] in is the speed of particle i in the k+1 update, is the speed of particle i in the kth update, θ is the particle speed adaptive inertia weight, q 1 ,q 2 is the learning factor, r 1 、r 2 is a random number from (0,1), is the position of particle i in the kth update, ζ is the regularization weight, is the position of particle i in the k+1th update, Levy is the Levy flight step length, θ max ,θ min is the maximum and minimum value of the adaptive inertia weight, K is the maximum number of updates, q 1s ,q 2s is the initial value of the learning factor, q 1e ,q 2e is the termination value of the learning factor, α, β, γ are dynamic adjustment coefficients;

[0111] Set the crossover threshold When the similarity between two bodies is greater than the crossover critical value, a crossover operation is performed:

[0112] x c =p mix x p1 +(1-p mix )x p2 +Gauss·σ

[0113]

[0114] where x c is the position of the descendant particle, p mix is the hybridization probability, x p1 、x p2 is the position of the particles of the two parents, Gauss is a Gaussian distribution random number, σ is the standard deviation of the Gaussian distribution, v c is the velocity of the descendant particle, v p1 、v p2 are the velocities of the two parent particles;

[0115] Use dynamic Cauchy mutation strategy to perform mutation operation:

[0116]

[0117] g best =p best (1+0.5Cauchy)

[0118] Where Cauchy is the Cauchy mutation strategy, p var is the adaptive mutation probability, ε is the dynamic adjustment coefficient;

[0119] Iteratively update the particle position and speed until the power storage and distribution objective function is minimized and then stop updating to obtain the optimal position g of the historical population. best And output the optimized power distribution plan, and upload the optimized power distribution plan and the corresponding first photovoltaic power generation group feature vector to the photovoltaic power generation history database;

[0120] In the actual evaluation, the predicted photovoltaic power generation is 30kW, which is less than the current user power supply demand of 50kW, and the calculated power ratio is 60%. At this time, it is necessary to optimize the current power distribution plan, discharge the power storage equipment, and provide the remaining power demand of each power supply area by the power grid;

[0121] Match the feature vector of the first photovoltaic power generation group with the photovoltaic power generation history database to obtain the historical power distribution plan: area A (photovoltaic modules 1, 2, 3, 4, 5 provide 18.75kW), area B (photovoltaic modules 6, 7 provide 7.5kW), area C (photovoltaic module 8 provides 3.75kW);

[0122] The particle velocity adaptive inertia weight θ = 0.9, the maximum and minimum values ​​of the adaptive inertia weight θ max =0.9,θ min =0.4, maximum number of updates K = 100, initial value of learning factor q 1s =q 2s =2, learning factor termination value q 1e =q 2e =0.5, dynamic adjustment coefficient α=β=γ=0.1, regularization weight ζ=0.01;

[0123] Adaptive mutation probability p var The expression is:

[0124]

[0125] where p var,max =0.1 is the maximum mutation probability, p var,min =0.01 is the minimum mutation probability, f max is the maximum fitness of the population, f avg is the average population fitness, f is the current population fitness, is the dynamic adjustment coefficient;

[0126] The optimized power distribution plan is continuously updated, and the weight coefficients of areas A, B, and C are 0.4, 0.3, and 0.3 respectively, and the change rate weight w 2 =0.2, vector weight w 21 =0.3, parameter weight w 22 =0.2, power distribution cost weight w 3 =0.3, the number of distribution lines m = 8, the unit distance cost c tran=0.05 yuan / km, the distances of the distribution lines are 3, 5, 4, 6, 7, 8, 9, 10 km, the transmission powers are 8, 6, 5, 6, 7, 5.5, 6.5, 7.5 kW, and the resistivities are 0.02, 0.03, 0.025, 0.028, 0.032, 0.026, 0.029, 0.031;

[0127] The calculated power distribution cost is 1.48695, the power supply satisfaction is 0.1533, the change rate is 0.012, and the energy storage cost is 0.072. At this time, the minimum power storage and distribution objective function is 1.7243, and the corresponding power storage and distribution schemes are area A (PV panels 7, 8, and 9 provide 11.25kW), area B (PV panels 1, 2, 3, and 4 provide 15kW), and area C (PV panel 8 provides 3.75kW).

[0128] In a second aspect, a photovoltaic power generation management system based on multi-source data includes:

[0129] Data acquisition module: used to obtain the operation data and status data of the photovoltaic power generation group, and pre-process the operation data and the status data;

[0130] Equipment parameter module: used for classifying and screening the operation data and the equipment status data to obtain photovoltaic power generation abnormal data, and determining the photovoltaic power generation group equipment parameter management plan according to the photovoltaic power generation abnormal data;

[0131] Photovoltaic power generation prediction module: used to construct a photovoltaic power generation prediction model according to the update data and the environmental status data, and input the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation;

[0132] Power storage and distribution scheme module: used to construct a power storage and distribution objective function, and determine the power storage and distribution scheme of the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function;

[0133] Management module: used to store, view and manage the operation data, the status data, the abnormal photovoltaic power generation data, the equipment parameter management plan and the power storage and distribution plan, perform abnormality detection and early warning based on the abnormal photovoltaic power generation data, and perform photovoltaic power generation management based on the equipment parameter management plan and the power storage and distribution plan.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic power generation management method based on multi-source data, characterized in that: The following steps are involved: S1. Acquire operation data and status data of a photovoltaic power generation group, and pre-process the operation data and the status data; the status data includes environmental status data and equipment status data; S2, classifying and screening the operation data and the equipment status data to obtain photovoltaic power generation abnormality data, determining a photovoltaic power generation group equipment parameter management plan according to the photovoltaic power generation abnormality data, and adjusting the photovoltaic power generation group equipment parameters; S3, obtaining updated data of the photovoltaic power generation group, constructing a photovoltaic power generation prediction model according to the updated data and the environmental status data, and inputting the operating data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain predicted photovoltaic power generation; S4. Construct a power storage and distribution objective function, determine a power storage and distribution plan for the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function, and manage photovoltaic power generation according to the equipment parameter management plan and the power storage and distribution plan.

2. The photovoltaic power generation management method based on multi-source data according to claim 1, characterized in that: The method for obtaining the photovoltaic power generation abnormal data comprises: Obtaining operation data and status data of a photovoltaic power generation group; the photovoltaic power generation group includes photovoltaic modules, inverters, energy storage equipment and power distribution equipment; the status data includes environmental status data and equipment status data; the environmental status data is the current meteorological information and meteorological environment information forecast provided by the Meteorological Bureau; A hierarchical clustering method is used to classify the operation data and equipment status data of the photovoltaic power generation group to obtain operation indicators and equipment status indicators; the operation indicators include photovoltaic component operation parameters, inverter operation parameters, energy storage device operation parameters and distribution device operation parameters; the equipment status indicators include photovoltaic component status parameters, inverter status parameters, energy storage device status parameters and distribution device status parameters; Calculating the equipment energy efficiency index of the photovoltaic power generation group according to the operation index and the equipment status index; the equipment energy efficiency index includes photovoltaic module energy efficiency parameters, inverter energy efficiency parameters, energy storage equipment energy efficiency parameters and power distribution energy efficiency parameters; According to the hierarchical clustering results, multiple hyperspheres are constructed to cover different distribution areas of normal data corresponding to operation indicators, equipment status indicators and equipment energy efficiency indicators, and abnormal screening is performed to obtain abnormal data of photovoltaic power generation.

3. The photovoltaic power generation management method based on multi-source data according to claim 1, characterized in that: The method for determining the photovoltaic power generation group equipment parameter management scheme includes: When the PV module status parameters are abnormal, the corresponding PV module status parameters will be adjusted according to the current meteorological information. When the PV module operation parameters and PV module energy efficiency parameters are abnormal, the PV module status parameters will be adjusted, and the shadow shielding, wiring conditions and PV panel working hours will be checked, and surface dust cleaning and heat dissipation operations will be performed; When the inverter operating parameters and inverter status parameters are abnormal, optimize the inverter control strategy and heat dissipation strategy; when the inverter energy efficiency parameters are abnormal, repair or replace the inverter; When the state parameters of the energy storage equipment are abnormal, adjust the corresponding state parameters of the energy storage equipment according to the operating parameters of the power distribution equipment; when the operating parameters of the energy storage equipment are abnormal, optimize the state parameters of the energy storage equipment; when the operating parameters of the energy storage equipment and the energy efficiency parameters of the energy storage equipment are abnormal, check or replace old batteries; When the distribution equipment operates abnormally, check the faults of photovoltaic modules and inverters. When the distribution equipment operating parameters, distribution equipment status parameters and distribution equipment energy efficiency parameters are abnormal, optimize the storage and distribution plan.

4. The photovoltaic power generation management method based on multi-source data according to claim 1, characterized in that: The method for obtaining the predicted photovoltaic power generation comprises: The photovoltaic power generation group update data and the current meteorological information are combined into a comprehensive data set, and the comprehensive data set is divided into a training set and a test set according to an 8:2 ratio; the photovoltaic power generation group update data includes operation update data and equipment status update data; Constructing a photovoltaic power generation prediction model, wherein the photovoltaic power generation prediction model includes a data preprocessing layer, a multimodal prediction layer, a strategy layer, and an output layer; The data preprocessing layer performs spatiotemporal alignment of the input data according to the time information and geographic information, and performs standardization and feature construction on the spatiotemporal aligned data to output the corresponding feature vector; The multimodal prediction layer predicts the corresponding photovoltaic power generation according to the equipment status and meteorological information of the photovoltaic power generation group, including a random forest base model, a long short-term memory network base model and a support vector machine base model; the random forest base model fits the nonlinear relationship between the operation data, the equipment status data and the meteorological information, and predicts the photovoltaic power generation, and the mean square error loss function is used to evaluate the model performance; the long short-term memory network base model processes the input data of time continuity, captures the dependency between the data, predicts the change trend of the photovoltaic power generation in the time series, uses the mean square error loss function to calculate the square difference between the predicted value and the actual value, and uses the Adam optimizer to adjust the learning rate; the support vector machine base model maps the equipment status data and the meteorological data to the high-dimensional space through the kernel function, finds the optimal hyperplane to predict the photovoltaic power generation, uses the ε-insensitive loss function to adjust the loss according to the error between the predicted value and the actual value, and optimizes the model parameters by solving the convex optimization problem; The strategy layer uses stacking fusion to fuse the photovoltaic power generation prediction results of the three base models, outputs the predicted photovoltaic power generation through the output layer, and uses the test set data to verify the photovoltaic power generation prediction model; The operation data and status data of the photovoltaic power generation group to be predicted are input into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation.

5. The photovoltaic power generation management method based on multi-source data according to claim 1, characterized in that: The method for constructing the power storage and distribution objective function comprises: The storage and distribution objective function is constructed based on the user's power supply satisfaction, the overall change rate of the distribution plan, the distribution cost and the energy storage cost. The expression is: where F aim is the power storage and distribution objective function, w1 is the power supply satisfaction weight, n is the number of power supply areas, w 1i is the weight coefficient of region i, is the electricity demand of region i, is the photovoltaic power supply in region i, w2 is the change rate weight, w 21 is the vector weight, w 22 is the parameter weight, X new is the characteristic vector of the new power distribution scheme, X old is the feature vector of the original power distribution scheme, ‖·‖2 is the Euclidean distance of the feature vector, δ(x k ) is the characteristic parameter x of the power distribution scheme k The penalty value, w3 is the distribution cost weight, m ​​is the number of distribution lines, c tran is the unit distance cost, d j is the distance of distribution line j, P j is the transmission power of distribution line j, λ is the loss cost coefficient, R j is the resistivity of distribution line j, t j is the power supply duration of distribution line j, w4 is the energy storage cost weight, c cap is the unit capacity cost, E rate is the rated capacity of energy storage, N cycle is the nominal cycle number, η is the charge and discharge loss coefficient, is the energy storage power in the time period T The sum of the absolute values ​​of .

6. The photovoltaic power generation management method based on multi-source data according to claim 1, characterized in that: The method for determining the photovoltaic power generation group storage and distribution scheme comprises: The current user power supply demand is determined by the distribution equipment operation data, and the photovoltaic power generation group storage and distribution plan is determined by comparing the predicted photovoltaic power generation with the current user power supply demand: When the predicted photovoltaic power generation is greater than the current user power supply demand, power distribution continues according to the original power distribution plan, and the state parameters of the power storage equipment are adjusted according to the overflow power to store power; When the predicted photovoltaic power generation is less than the current user power supply demand, the power ratio of the predicted photovoltaic power generation to the current user power supply demand is calculated, the power distribution plan is selected according to the power ratio, and the state parameters of the power storage device are adjusted according to the power ratio for discharge; The specific method of selecting a power distribution scheme according to the power ratio is as follows: When the power ratio is less than or equal to 50%, all the predicted photovoltaic power generation and power storage equipment power will be allocated to the nearest power supply area, and the power demand of the remaining power supply area will be provided by the power grid; When the power ratio is greater than 50%, the current power distribution plan is optimized, and the remaining power demand of each power supply area is provided by the power grid; the power distribution plan includes the power supply areas allocated to different photovoltaic power generation groups and power storage devices, and the allocated power obtained by the power supply areas; The specific method for optimizing the current power distribution scheme is: The predicted photovoltaic power generation feature vector, the current user power demand feature vector, the photovoltaic power generation group state data feature vector and the current meteorological information feature vector are combined into a first photovoltaic power generation group feature vector, and the first photovoltaic power generation group feature vector is matched with a photovoltaic power generation history database to obtain a historical power distribution plan; the photovoltaic power generation history database includes historical photovoltaic power generation group feature vectors and corresponding power distribution plans; the method for obtaining the historical power distribution plan is to calculate the comprehensive similarity between the first photovoltaic power generation group feature vector and the historical photovoltaic power generation group feature vector, and take the power distribution plan corresponding to the historical photovoltaic power generation group feature vector with the highest comprehensive similarity as the historical power distribution plan; According to the power storage and distribution objective function, the historical power distribution plan is optimized, and the particle swarm algorithm is used to take the historical power distribution plan as the initial particle population to determine the current individual optimal position p best and the optimal position of the historical population g best , update the particle velocity and particle position: in is the speed of particle i in the k+1 update, is the speed of particle i in the kth update, θ is the particle speed adaptive inertia weight, q1 and q2 are learning factors, r1 and r2 are random numbers (0,1), is the position of particle i in the kth update, ζ is the regularization weight, is the position of particle i in the k+1th update, Levy is the Levy flight step length, θ max ,θ min is the maximum and minimum value of the adaptive inertia weight, K is the maximum number of updates, q 1s ,q 2s is the initial value of the learning factor, q 1e ,q 2e is the termination value of the learning factor, α, β, γ are dynamic adjustment coefficients; Set the crossover threshold When the similarity between two bodies is greater than the crossover critical value, a crossover operation is performed: x c =p mix x p1 +(1-p mix )x p2 +Gauss·s where x c is the position of the descendant particle, p mix is the hybridization probability, x p1 、x p2 is the position of the particles of the two parents, Gauss is a Gaussian distribution random number, σ is the standard deviation of the Gaussian distribution, v c is the velocity of the descendant particle, v p1 、v p2 are the velocities of the two parent particles; Use dynamic Cauchy mutation strategy to perform mutation operation: g best =p best ·(1+0.5Cauchy) Where Cauchy is the Cauchy mutation strategy, p var is the adaptive mutation probability, ε is the dynamic adjustment coefficient; Iteratively update the particle position and speed until the power storage and distribution objective function is minimized and then stop updating to obtain the optimal position g of the historical population. best The optimized power distribution plan is output, and the optimized power distribution plan and the corresponding first photovoltaic power generation group feature vector are uploaded to the photovoltaic power generation history database.

7. A photovoltaic power generation management system based on multi-source data, used to execute the method according to any one of claims 1 to 6, characterized in that: include: Data acquisition module: used to obtain the operation data and status data of the photovoltaic power generation group, and pre-process the operation data and the status data; Equipment parameter module: used for classifying and screening the operation data and the equipment status data to obtain photovoltaic power generation abnormal data, and determining the photovoltaic power generation group equipment parameter management plan according to the photovoltaic power generation abnormal data; Photovoltaic power generation prediction module: used to construct a photovoltaic power generation prediction model according to the update data and the environmental status data, and input the operation data and status data of the photovoltaic power generation group to be predicted into the photovoltaic power generation prediction model to obtain the predicted photovoltaic power generation; Power storage and distribution scheme module: used to construct a power storage and distribution objective function, and determine the power storage and distribution scheme of the photovoltaic power generation group according to the predicted photovoltaic power generation and the power storage and distribution objective function; Management module: used to store, view and manage the operation data, the status data, the equipment parameter management plan and the power storage and distribution plan, and perform photovoltaic power generation management according to the equipment parameter management plan and the power storage and distribution plan.

Images