Efficiency evaluation method of photovoltaic power generation cold storage system based on big data analysis

Abstract

The invention provides an efficiency evaluation method for a photovoltaic power generation cold storage system based on big data analysis, and the method comprises the steps: obtaining the shadow distribution data of the surface of a photovoltaic module, extracting the change characteristics of a shielding area through an image processing method, and judging the dynamic change trend of the shielding area in combination with a time sequence analysis method; according to the dynamic change trend of the shielding area, predicting the power generation efficiency fluctuation characteristics of the photovoltaic module by adopting a regression model, and establishing a mapping relation between the power generation efficiency and the shielding area; and the strategy adjustment scheme meeting the threshold condition is applied to the cold storage system, the change trend of the system efficiency is monitored in real time, a cold storage system efficiency evaluation report is generated in combination with a big data analysis method, and the evaluation report comprises the incidence relation among the shielding area, the power generation efficiency, the cooling capacity storage and the system efficiency.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an efficiency evaluation method for a photovoltaic power generation cold storage system based on big data analysis. Background Art

[0002] Covering the cold storage system with photovoltaic panels can generate electricity from solar energy and provide energy for the cold storage system. At the same time, the photovoltaic panels can also play a shading role, which helps to reduce the energy consumption of the cold storage system. However, in the urban building environment, due to the uneven floors of the surrounding buildings, the shadows of the surrounding buildings will cause different local shading of the photovoltaic components above the cold storage system as the direction of the sun changes, resulting in fluctuations in the power generation efficiency of the photovoltaic components. Since the shading area of ​​the photovoltaic components changes dynamically over time, this fluctuation will affect the output power of the photovoltaic power generation system, and then affect the operation of the cold storage system coupled with it. The output of the cold storage system based on photovoltaic power generation is closely related to the power generation efficiency of the photovoltaic components. Therefore, the cold storage strategy of the cold storage system needs to be dynamically adjusted according to the fluctuation of the power generation efficiency of the photovoltaic components. However, when the shading area changes dynamically, how to adjust the cold storage strategy of the cold storage system in real time and the mechanism of the impact of this adjustment on the overall system operation efficiency are still unclear. Specifically, when the photovoltaic components are partially shaded, their output power will decrease, resulting in a reduction in the available power energy of the cold storage system. At this time, the cold storage system needs to decide whether to reduce the cooling capacity to match the real-time output of photovoltaic power generation, or to maintain the original cooling capacity and make up for the power gap by consuming energy storage. This decision needs to take into account multiple factors such as the forecast of photovoltaic power generation, the cooling load demand of the building, the energy storage capacity and efficiency of the cold storage system. Different decision-making strategies will have different effects on the operating efficiency of the overall system, but their internal mechanisms need further study. How to dynamically optimize and adjust the cold storage strategy of the cold storage system in the case of fluctuations in power generation efficiency caused by partial shading of photovoltaic modules, so as to maximize the overall operating efficiency of the cold storage system while ensuring the cooling needs of the building, is a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides an efficiency evaluation method for a photovoltaic power generation cold storage system based on big data analysis, which mainly includes:

[0004] Obtain the shadow distribution data on the surface of the photovoltaic module, extract the change characteristics of the shielding area through image processing methods, and judge the dynamic change trend of the shielding area by combining the time series analysis method; according to the dynamic change trend of the shielding area, use the regression model to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic module, and establish a mapping relationship between the power generation efficiency and the shielding area; according to the fluctuation characteristics of the power generation efficiency, analyze the historical operation data and meteorological data of the photovoltaic power generation system, extract the change law of the cold storage demand of the cold storage system, and determine the adjustment content of the cold storage strategy. The adjustment content includes dynamic adjustment parameters, and the dynamic adjustment parameters include adjustment of cold storage time, cold storage power and cold storage temperature parameters; according to the dynamic adjustment parameters of cold storage, combined with the performance parameters and operation constraints of the cold storage system, analyze the impact of the cold storage adjustment plan on the operation of the cold storage system through energy efficiency analysis methods. The influencing mechanism of efficiency includes the influence of the adjustment plan on the cold storage capacity, cold release capacity and energy consumption of the cold storage system, and the change characteristics of the system efficiency are obtained. The change characteristics include the system comprehensive energy efficiency, the operating cost under the peak and valley electricity prices, and the carbon emission intensity. Through the change characteristics of the system efficiency, it is judged whether the operating status of the cold storage system meets the preset threshold conditions. If the improvement of the system comprehensive energy efficiency exceeds the preset threshold, the reduction of the operating cost exceeds the preset threshold, and the reduction of the carbon emission intensity exceeds the preset threshold, it is judged that the threshold conditions are met, and the current strategy adjustment plan is output. The strategy adjustment plan that meets the threshold conditions is applied to the cold storage system, and the changing trend of the system efficiency is monitored in real time. Combined with the big data analysis method, an efficiency evaluation report of the cold storage system is generated. The evaluation report includes the correlation between the shielding area, power generation efficiency, cold storage and system efficiency.

[0005] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0006] The present invention discloses an efficiency evaluation method for a photovoltaic power generation cold storage system based on big data analysis. The method analyzes the dynamic changes of the shadow distribution on the surface of photovoltaic modules, establishes a mapping relationship between the shielding area and the power generation efficiency, and predicts the fluctuation characteristics of the power generation efficiency. Combined with historical operation and meteorological data, the changing law of the cold storage demand of the cold storage system is extracted, and the dynamic adjustment parameters of the cold storage strategy are determined. Through energy efficiency analysis, the impact of strategy adjustment on the system operation efficiency is evaluated, including comprehensive energy efficiency, operation cost and carbon emission intensity. When the preset threshold conditions are met, the optimization strategy is applied and the system efficiency is monitored in real time. The present invention realizes the coordinated optimization of photovoltaic power generation and cold storage systems, improves the overall energy utilization efficiency, reduces the operation cost and carbon emissions, and provides a new technical solution for the integrated application of renewable energy and building energy conservation. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 The present invention is a flow chart of a method for evaluating the efficiency of a photovoltaic power generation cold storage system based on big data analysis.

[0008] Figure 2 It is a schematic diagram of an efficiency evaluation method of a photovoltaic power generation cold storage system based on big data analysis according to the present invention.

[0009] Figure 3 It is another schematic diagram of the efficiency evaluation method of a photovoltaic power generation cold storage system based on big data analysis according to the present invention. DETAILED DESCRIPTION

[0010] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] like Figure 1-3 In this embodiment, a method for evaluating the efficiency of a photovoltaic power generation cold storage system based on big data analysis may specifically include:

[0012] Step S101, obtaining shadow distribution data on the surface of the photovoltaic module, extracting the change characteristics of the shielding area through image processing methods, and combining with time series analysis methods to determine the dynamic change trend of the shielding area.

[0013] An image acquisition device is used to obtain a high-resolution image of the surface of a photovoltaic module, and a shadow area is divided according to a comparison result between the grayscale value of a pixel point of the high-resolution image and a preset reference grayscale value to obtain a first shadow area value; edge detection is performed on the shadow area to obtain an edge contour line, and the edge contour line is fitted by the least squares method to obtain an edge contour curve equation, and an autoregressive moving average model is established according to the edge contour curve equation to obtain a shadow area motion trajectory equation; temperature data is collected by a pre-deployed temperature sensor array to construct a temperature distribution matrix, and isotherms are extracted for the temperature distribution matrix. If the temperature value in the isotherm closed curve is lower than the temperature value of the surrounding area, it is determined that the area is shadowed, and a second shadow area value is obtained; the first shadow area value and the second shadow area value are fused by a weighted average method, and a discrete Fourier transform is performed on the fused shadow area value sequence, and frequency components with amplitudes greater than a normalized threshold are extracted according to the spectrum diagram of the discrete Fourier transform to obtain a shadow area change periodic sequence.

[0014] Exemplarily, a high-resolution image is collected from the surface of a photovoltaic module, and the collected image is grayed. The shadow area and the non-shadow area are divided according to the comparison of the gray value of the pixel point with the preset reference gray value, and the number of pixels in the shadow area is counted to obtain the first shadow area value. The shadow area is edge detected, the edge contour line is extracted, the edge contour curve equation is fitted by the least squares method, and the difference of the edge contour curve equation coefficients at two adjacent sampling moments is calculated. According to the difference sequence of the edge contour curve equation coefficients, an autoregressive moving average model is established, the moving speed vector of the shadow area is calculated, and the motion trajectory equation of the shadow area on the surface of the photovoltaic module is obtained. The temperature data is collected by using the temperature sensor array pre-arranged on the surface of the photovoltaic module, and a temperature distribution matrix is ​​constructed. The temperature distribution matrix is ​​interpolated in two dimensions to obtain a continuous temperature distribution function. The temperature distribution function is subjected to isotherm extraction. If the temperature value within the isotherm closed curve is lower than the temperature value of the surrounding area, it is determined that there is a shadow shielding within the closed curve range, and the second shadow area value is calculated. The weighted average method is used to fuse the first shadow area value and the second shadow area value to obtain a comprehensive shadow area value sequence, and the sequence is subjected to discrete Fourier transform. The frequency components with amplitudes greater than the normalized threshold are extracted from the spectrum after Fourier transformation, and the periodic values ​​corresponding to each frequency component are calculated to generate a periodic sequence of shadow area changes. When collecting grayscale images of the photovoltaic module surface, an industrial camera with a resolution of 2048×2048 pixels is used to shoot under natural lighting conditions. The grayscale value of each pixel in the image ranges from 0 to 255, where 0 represents full black and 255 represents full white. In practical applications, the grayscale value of the shadow area is usually lower than 100, and the grayscale value of the non-shadow area is usually higher than 200. Therefore, 150 is selected as the preset reference grayscale value to distinguish between shadow and non-shadow areas. Shadow edge detection is processed using the Sobel operator. The edge contour point set is obtained by calculating the grayscale gradient of the image in the horizontal and vertical directions. The coordinate accuracy of each edge point can reach 0.1 pixel. For a photovoltaic module with an area of ​​1 square meter, 300 to 500 edge contour points can usually be extracted, and the distribution of these points can better reflect the shape characteristics of the shadow area. In actual application scenarios, the shadows on the surface of photovoltaic modules mainly come from surrounding buildings, trees and other obstructions, and the movement of these obstructions has a certain regularity. Taking the shadow caused by the swing of branches as an example, when the wind speed is 3 meters per second, the swing period of the branches is about 2 seconds, and the displacement speed of the corresponding shadow edge contour points varies between 0.1 and 0.3 meters per second. The surface temperature distribution of photovoltaic modules is measured using an 8×8 array thermocouple sensor with a sensor spacing of 0.1 meters and a measurement accuracy of 0.1 degrees Celsius. Under standard test conditions, the temperature in the non-shadow area is usually between 45 and 50 degrees Celsius, while the temperature in the shadow area will drop by 5 to 8 degrees Celsius due to the weakening of light intensity. By performing bicubic spline interpolation on the temperature data, a continuous temperature distribution function with a resolution of 0.01 meters can be obtained.In the periodic analysis of the shadow area numerical series, for data with a sampling frequency of 10 Hz and a duration of 60 seconds, a frequency component of 0.02 Hz can be distinguished after a 512-point fast Fourier transform. In practical applications, the main period of shadow area change is usually between 1 and 10 seconds, and the corresponding frequency component amplitude is significantly higher than other frequency components. By setting the normalized amplitude threshold to 0.1, meaningful periodic components can be effectively extracted. When calculating the comprehensive shadow area, considering the characteristics of the two methods based on grayscale value and temperature value, weights of 0.7 and 0.3 are assigned respectively. This is because the grayscale image has a higher spatial resolution and can more accurately reflect the shadow boundary, while the temperature data, although with a lower resolution, is not affected by ambient light and can serve as an effective supplement.

[0015] Step S102, according to the dynamic change trend of the shielding area, a regression model is used to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic module, and a mapping relationship between the power generation efficiency and the shielding area is established.

[0016] The fluctuation component in the power generation efficiency sequence is extracted through a sliding time window to obtain a first power generation efficiency curve; a nonlinear mapping function is constructed according to the first power generation efficiency curve by using a support vector regression method to obtain a first prediction function; for the first prediction function, a temperature correction coefficient is calculated according to a temperature coefficient to obtain a second prediction function; the second prediction function is used to process the real-time shielding area data to obtain a second power generation efficiency curve, and the high-frequency component and the low-frequency component in the second power generation efficiency curve are extracted by wavelet transform to obtain a third power generation efficiency curve; the third power generation efficiency curve is subjected to Fourier transform to extract the spectrum feature vector, and the Bayesian regression method is used to fuse the environmental temperature and humidity data to predict the future power generation efficiency change trend curve.

[0017] Exemplarily, a string photovoltaic module is used to collect voltage and current data, and the power generation value is calculated based on the voltage and current data. The first power generation efficiency value sequence is obtained by combining the standard light intensity of 1000 watts / square meter. The fluctuation component in the power generation efficiency sequence is extracted through a sliding time window of 60 seconds to obtain the first power generation efficiency fluctuation curve. For the shading area data sequence and the first power generation efficiency fluctuation curve, the support vector regression method of the Gaussian kernel function is used to construct a nonlinear mapping function, and the corresponding relationship between the shading area change and the power generation efficiency fluctuation amplitude is extracted to obtain the first prediction function. The deviation value between the temperature mean and the standard temperature of 25 degrees Celsius is extracted from the surface temperature distribution data of the photovoltaic module, and the temperature correction coefficient is calculated according to the temperature coefficient minus 0.4% / degree Celsius. The first prediction function is corrected to obtain the second prediction function. The second prediction function is used to process the real-time shading area data to obtain the second power generation efficiency fluctuation curve, and the high-frequency component and low-frequency component in the fluctuation curve are extracted by wavelet transform. According to the ratio of the rated photoelectric conversion efficiency value on the nameplate of the photovoltaic module to the measured photoelectric conversion efficiency value, the efficiency attenuation coefficient is calculated, and the high-frequency component and low-frequency component of the second power generation efficiency fluctuation curve are combined to obtain the third power generation efficiency fluctuation curve. The third power generation efficiency fluctuation curve is subjected to Fourier transform, the spectrum feature vector is extracted, and the Bayesian regression method is used to fuse the ambient temperature and humidity data to predict the future power generation efficiency change trend curve. By calculating the standard deviation of the power generation efficiency change trend curve, if the standard deviation exceeds the preset fluctuation threshold of 0.05, the power generation efficiency fluctuation is determined to be abnormal. The power generation efficiency of photovoltaic modules is closely related to environmental conditions, among which the shading of the module surface has a significant impact on the power generation efficiency. In actual applications, a typical 60-cell module has an output voltage of 36 volts and an output current of 9 amperes under standard lighting conditions, corresponding to a power generation power of 324 watts. When the module is partially shielded, the output power of the shielded cell drops sharply, causing the power generation efficiency of the entire string of modules to fluctuate. Changes in the shielding area on the surface of the module will cause the power generation efficiency to present different fluctuation characteristics. Taking the dynamic shading caused by tree branch projection as an example, when the shading area increases from 5% to 10%, the power generation efficiency usually decreases by 8% to 12%. The power generation efficiency data is processed with a sliding time window of 60 seconds, which can effectively capture this fluctuation feature. Temperature also has an important impact on the power generation efficiency of photovoltaic modules. For every 1 degree Celsius increase in the surface temperature of the module, the power generation efficiency decreases by about 0.4%. The measured data shows that the module conversion efficiency is 18% at 25 degrees Celsius, and when the temperature rises to 45 degrees Celsius, the conversion efficiency drops to 16.6%. By introducing a temperature correction coefficient to correct the prediction model, the accuracy of the power generation efficiency prediction is improved. The power generation efficiency fluctuation of photovoltaic modules has multi-scale characteristics. The fluctuation can be decomposed into different frequency components through wavelet transform: the high-frequency component reflects instantaneous disturbances, such as short-term light changes caused by cloud movement; the low-frequency component reflects the long-term change trend, such as the efficiency drift caused by temperature changes.The measured data show that the high-frequency fluctuation amplitude does not exceed 3% under normal working conditions, and the low-frequency fluctuation amplitude does not exceed 5%. The photoelectric conversion efficiency gradually decreases with the service life of the component. The conversion efficiency of the newly installed component is 19.5%, which drops to 19% after one year of use and to 18% after five years. The efficiency attenuation coefficient reflects the aging degree of the component, which is closely related to the fluctuation characteristics of the power generation efficiency. The more severe the attenuation, the greater the efficiency fluctuation caused by the same change in the shielding area. The spectral characteristics of the power generation efficiency fluctuation contain rich diagnostic information. The Fourier transform of the fluctuation curve shows that the abnormal spectral components appearing in the 0.1 to 1 Hz frequency band often indicate that the component has a shielding problem. The actual operation data shows that the standard deviation of the power generation efficiency fluctuation under normal working conditions is usually less than 0.05. When this threshold is exceeded, the shielding of the component needs to be further investigated. Under the standard test conditions of an ambient temperature of 20 degrees Celsius and a relative humidity of 60%, the power generation efficiency fluctuation of the photovoltaic module shows a relatively stable characteristic. Through the Bayesian regression method, the influence of temperature and humidity changes on the power generation efficiency is comprehensively considered, and the average deviation between the predicted results and the measured data is controlled within 2%.

[0018] According to the dynamic change trend of the shading area, a regression model with the shading area as the independent variable and the power generation efficiency as the dependent variable is established to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic modules as the shading area changes. The quantitative mapping relationship between the power generation efficiency and the shading area is fitted, including the change range of the shading area, the fluctuation amplitude of the power generation efficiency and the peak efficiency point, so as to predict the power generation efficiency change curve under different shading scenarios.

[0019] The light intensity data is collected according to the light intensity sensor array, a light intensity distribution matrix is ​​constructed for the light intensity data, and a threshold segmentation method is used to process the light intensity distribution matrix to obtain a binary image of the occluded area; the pixel ratios of the occluded area in the binary image of the occluded area are counted to obtain a first numerical sequence of occluded area, and the first numerical sequence of occluded area is subjected to Fourier transformation to extract spectral features, and a trigonometric function superposition fitting is used to obtain a dynamic change function of the occluded area; the voltage and current sampling data output by the photovoltaic module are obtained, the power generation value is calculated according to the voltage and current sampling data, and a fixed time window is used to perform sliding average processing on the power generation value to obtain a first power generation efficiency curve; the sampling value of the dynamic change function of the occluded area is used as an independent variable, and the corresponding time value of the first power generation efficiency curve is used as a dependent variable, and a first prediction function is obtained by training with a support vector regression method.

[0020] Exemplarily, the light intensity data is collected by the light intensity sensor array pre-arranged on the surface of the photovoltaic module, and the light intensity distribution matrix is ​​constructed according to the uniform distribution position of the light intensity sensor array on the horizontal plane. The threshold segmentation is applied to the light intensity distribution matrix to obtain a binary image of the occluded area, and the pixel ratio of the occluded area in the binary image is statistically obtained to obtain a first numerical sequence of occluded area. The first numerical sequence of occluded area is subjected to fast Fourier transform, and the first three frequency components with the largest amplitude in the spectrum characteristics are extracted. The dynamic change function of the occluded area is obtained by superposition fitting of trigonometric functions. The power generation value is calculated according to the voltage and current sampling data output by the photovoltaic module, and the power generation value is sliding averaged with a fixed time window length to obtain the first power generation efficiency curve after smoothing. The sampling value of the dynamic change function of the occluded area is used as the independent variable, and the corresponding time value of the first power generation efficiency curve is used as the dependent variable. The regression function is trained by the support vector regression method to obtain the first regression prediction function. The first regression prediction function is numerically derived, and the dividing point between the increasing interval and the decreasing interval of the function value is extracted, and the maximum and minimum points of the power generation efficiency are determined to obtain the boundary value of the power generation efficiency fluctuation range. The shading working condition is divided according to the boundary value of the power generation efficiency fluctuation range. The power generation efficiency value is supplemented by cubic spline interpolation in each working condition interval to obtain the second power generation efficiency curve. The second power generation efficiency curve is used as training data, and the gradient boosting tree algorithm is used to construct the second regression prediction function. The corresponding power generation efficiency output value is calculated according to the real-time shading area input value. The light intensity sensor array on the surface of the photovoltaic module is usually arranged in an 8×8 uniform manner, with a sensor spacing of 125 mm, covering the entire module surface. Each sensor collects the light intensity value in real time. The normal value of the light intensity is between 800 and 1000 watts / square meter. When shading occurs, the light intensity of the shaded area drops below 200 watts / square meter. By setting a threshold of 400 watts / square meter for image segmentation, a binary image reflecting the distribution of the shading area is obtained. The dynamic change of the shading area has obvious periodic characteristics. Fourier analysis of the shading area data with a sampling frequency of 10 Hz and a sampling time of 300 seconds found that the main frequency components were concentrated near 0.2, 0.4 and 0.8 Hz, corresponding to shading from different sources such as tree branches swinging and cloud movement. The fitting error between the shading area change function obtained by trigonometric function superposition fitting and the measured data is less than 3%. The power generation efficiency of photovoltaic modules is very sensitive to shading. Under standard test conditions, the rated power of a 60-cell module is 330 watts, and the corresponding power generation efficiency is 19.5%. Smoothing the power generation data with a sliding time window of 30 seconds can effectively eliminate the influence of instantaneous fluctuations. The measured data show that when the shading area increases from 0 to 5%, the power generation efficiency decreases by about 15%. The support vector regression method uses the radial basis kernel function to construct a nonlinear mapping relationship. 1000 sets of corresponding data of shading area and power generation efficiency were selected as training samples, and the root mean square error of cross-validation was 0.8%.The derivative of the regression function reflects the rate of change of power generation efficiency with the shading area. An inflection point appears at a shading area of ​​3%, corresponding to the minimum value of power generation efficiency. The fluctuation range of power generation efficiency varies with the shading conditions. Statistical analysis shows that under mild shading conditions (shading area less than 5%), the power generation efficiency fluctuates from 17% to 19%; under moderate shading conditions (shading area 5% to 10%), the fluctuation range expands to 15% to 18%; under severe shading conditions (shading area greater than 10%), the fluctuation range further expands to 12% to 17%. The training of the gradient boosting tree algorithm uses 500 decision trees, and the maximum depth of each tree is 4. The input features include shading area, shading change rate and shading duration, and the output is the predicted power generation efficiency value. The prediction accuracy under different shading conditions exceeds 95%, and the width of the confidence interval of the prediction results increases with the increase of shading area, reflecting that the uncertainty of the prediction is positively correlated with the degree of shading.

[0021] Step S103, analyzing the historical operation data and meteorological data of the photovoltaic power generation system in view of the fluctuation characteristics of power generation efficiency, extracting the changing law of the cold storage demand of the cold storage system, and determining the adjustment content of the cold storage strategy, the adjustment content includes dynamic adjustment parameters, and the dynamic adjustment parameters include adjusting the cold storage time, cold storage power and cold storage temperature parameters.

[0022] According to the photovoltaic power sampling data, the fast Fourier transform method is used to process the sampling data to obtain the periodic characteristics and amplitude characteristics of the power generation fluctuation; the periodic characteristics and amplitude characteristics are segmented, and the feature extraction method is used to obtain the meteorological feature vectors of light intensity, ambient temperature and relative humidity from the segmented data to obtain a feature sequence; according to the feature sequence, the support vector machine clustering method is used to process the feature sequence to obtain a cold storage load prediction curve; for the cold storage load prediction curve, a linear programming solver is used to calculate the start and end points of the cold storage time period to obtain a cold storage duration curve; the cold storage duration curve is segmentedly integrated to calculate the cold storage power demand in each cold storage time period.

[0023] Exemplarily, a time series is constructed based on the photovoltaic power sampling data, and the periodic characteristics and amplitude characteristics of the power generation fluctuation are extracted by the fast Fourier transform method. Combined with the temperature change curve data, the power generation fluctuation trend is smoothed by the Kalman filter to obtain the first power fluctuation curve and the fluctuation period value. The first power fluctuation curve is segmented according to the fluctuation period value, and the meteorological feature vectors of the three dimensions of light intensity, ambient temperature, and relative humidity are extracted in each time period to obtain the first feature sequence. The first feature sequence is clustered by a support vector machine, and the mapping relationship between the meteorological feature combination and the cold storage load is extracted from the clustering result, and the first cold storage load prediction curve is calculated. According to the first cold storage load prediction curve, the enthalpy balance equation group of the cold storage device is constructed, and the linear programming solver is used to calculate the start and end points of the cold storage time period to obtain the first cold storage duration curve. The first cold storage duration curve is segmented and integrated, and the cold storage power demand in each cold storage time period is calculated to obtain the first cold storage power curve. The kernel density estimation method is used to process the inlet and outlet water temperature data of the cold storage device, determine the peak point and trough point of the kernel density curve, and obtain the upper and lower limits of the cold storage temperature. According to the first cold storage load prediction curve, the first cold storage duration curve, and the first cold storage power curve, the weighted average method is used to update the cold storage control parameter value. The fluctuation of photovoltaic power generation has obvious time series characteristics. When the sampling frequency is 1 Hz, the collected power data shows multi-scale fluctuation characteristics: the short-term fluctuation period is between 60 seconds and 300 seconds, which is mainly affected by cloud movement; the medium-term fluctuation period is between 1800 seconds and 3600 seconds, which is mainly affected by ambient temperature changes. After Kalman filtering, the noise level of the power fluctuation curve is reduced by 90%. Meteorological characteristics show a significant correlation with cold storage load. Under typical summer conditions, when the ambient temperature rises from 25 degrees Celsius to 35 degrees Celsius, the cold storage load increases by about 40%. For every 10% increase in relative humidity, the cold storage load increases by about 5%. When the light intensity exceeds 800 watts / square meter, the growth rate of the cold storage load increases significantly. The key to determining the cold storage time lies in the enthalpy balance calculation. The inlet water temperature of the cold water energy storage device is usually controlled between 5 and 7 degrees Celsius, and the outlet water temperature is controlled between 12 and 15 degrees Celsius. When the capacity of the cold storage device is 1000 kilowatts, the time required to complete a complete cold storage cycle is between 4 and 6 hours. Through linear programming optimization calculation, the cold storage process can be divided into 3 to 5 time periods. The cold storage power demand varies significantly with the seasons. Under typical summer conditions, the cold storage power demand fluctuates between 800 and 1000 kilowatts during the daytime and drops to 400 to 600 kilowatts during the nighttime. Through segmented integral calculation, the cumulative cooling capacity in each cold storage time period is obtained, and then the average cold storage power of the period is determined. The control of the cold storage temperature has a significant impact on the system performance. The kernel density estimation results show that the optimal control range of the inlet water temperature is 6.5±0.5 degrees Celsius, and the optimal control range of the outlet water temperature is 13.5±0.5 degrees Celsius.For every 0.1 degree Celsius increase in temperature control accuracy, the cold storage efficiency increases by about 1%. The dynamic adjustment of cold storage control parameters needs to take into account multiple factors. Taking a 500-kilowatt cold storage device as an example, the cold storage time parameter value under summer conditions is 5 hours, the cold storage power parameter value is 850 kilowatts, and the cold storage temperature parameter value is 7 degrees Celsius. When the fluctuation amplitude of photovoltaic power generation exceeds 20%, the cold storage time is extended to 6 hours, the cold storage power is reduced to 700 kilowatts, and the cold storage temperature is increased to 8 degrees Celsius. The parameter update adopts the weighted average method, in which the weight of the cold storage load prediction value is 0.5, the weight of the cold storage time is 0.3, and the weight of the cold storage power is 0.2. By dynamically adjusting the weight coefficient, the timeliness of the system response is guaranteed, and the frequent fluctuations of the parameters are avoided. Actual operation data show that this method improves the operating efficiency of the cold storage system by more than 15%.

[0024] Mining the changing patterns of cold storage system's cooling demand over time and weather in historical operation and meteorological data, identifying the key parameters that affect cold storage demand, extracting the intraday fluctuations and seasonal trends of load demand, and determining the optimization space of cold storage strategy based on the changes in cooling demand, adjusting the action space of cold storage operation parameters, including the adjustable time period range, upper and lower power limits, and temperature setting allowable range, to form the boundary of dynamic optimization of cold storage strategy parameters.

[0025] According to the cooling load sequence and the meteorological parameter sequence, the long short-term memory network is used to extract the daily fluctuation law and seasonal change characteristics of the cooling load to obtain the first load fluctuation curve; the first load fluctuation curve is processed by wavelet transform, and the second load fluctuation curve is obtained by extracting the high-frequency fluctuation component and the low-frequency change trend; the second load fluctuation curve and the meteorological parameter sequence are processed by a decision tree regressor, and the load influence parameter sequence is obtained by calculating the influence of temperature, humidity and light intensity on the load change; the operation data of the cold storage device is classified according to the load influence parameter sequence, and the boundary points of the cold storage period are identified by the hierarchical clustering method of Euclidean distance to obtain the cold storage operation segmentation result; the inlet and outlet water temperature data sequence in each cold storage operation segment is Fourier transformed, and the temperature fluctuation range corresponding to the main frequency component is extracted to obtain the upper and lower limits of the cold storage temperature; the thermal enthalpy difference is calculated according to the upper and lower limits of the cold storage temperature, and the cold storage power adjustment range is obtained in combination with the rated capacity of the cold storage device.

[0026] Exemplarily, based on the cooling load sequence and meteorological parameter sequence in the historical operation data, the long short-term memory network is used to extract the intraday fluctuation law and seasonal change characteristics of the cooling load, and the load fluctuation mean and variance are calculated through the hourly sliding time window to obtain the first load fluctuation curve. Wavelet transform is applied to the first load fluctuation curve to extract the high-frequency fluctuation component and the low-frequency change trend to obtain the second load fluctuation curve. For the second load fluctuation curve and the meteorological parameter sequence, a decision tree regressor is constructed to extract the influence of temperature, humidity, and light intensity on the load change, and the weight coefficient of each parameter is calculated to obtain the load influence parameter sequence. The operation data of the cold storage device is classified according to the load influence parameter sequence, and the boundary points of the cold storage period are identified by the hierarchical clustering method of Euclidean distance to obtain the cold storage operation segmentation result. The inlet and outlet water temperature data sequence in each cold storage operation segment is Fourier transformed to extract the temperature fluctuation range corresponding to the main frequency component to obtain the upper and lower limits of the cold storage temperature. The thermal enthalpy difference is calculated according to the upper and lower limits of the cold storage temperature, and the cold storage power adjustment range is obtained in combination with the rated capacity of the cold storage device. The kernel density estimation method is used to process the cold storage data of the cold storage device. The cold storage density distribution curve is extracted by fixing the bandwidth parameter to determine the boundary value of the cold storage capacity. According to the cold storage operation segmentation results, the upper and lower limits of the cold storage temperature, the cold storage power adjustment interval and the boundary value of the cold storage capacity, the cold storage strategy parameter constraint matrix is ​​constructed. The cooling load usually shows typical intraday fluctuations and seasonal changes. Under summer conditions, the hourly sampled cooling load data show that the load peak occurs from 14:00 to 16:00, the load valley occurs from 4:00 to 6:00, and the peak-to-valley ratio reaches 2.5. The load fluctuation mean processed by a 3-hour sliding window varies between 500 kW and 800 kW, and the variance value varies between 50 kW and 100 kW. Wavelet transform can effectively separate the high-frequency fluctuations and low-frequency trends of the load curve. The high-frequency component reflects short-term load fluctuations, with a fluctuation period of 1 hour to 4 hours, and the fluctuation amplitude accounts for 15% to 25% of the total load. The low-frequency component reflects seasonal changes. From spring to summer, the average daily load increases by 40%, and from summer to autumn, the average daily load decreases by 35%. The degree of influence of meteorological parameters on cooling load varies. Decision tree regression analysis shows that for every 1 degree Celsius increase in outdoor temperature, the cooling load increases by about 8%; for every 10% increase in relative humidity, the cooling load increases by about 3%; when the sunshine intensity exceeds 600 watts / square meter, the cooling load growth rate increases significantly. The influence weight ratio of temperature, humidity, and light intensity is about 5:2:3. The division of cold storage operation time periods is based on load characteristics and environmental conditions. Using the hierarchical clustering method of Euclidean distance, the 24 hours of the day are divided into 4 operation periods: the late night cold storage period from 23:00 to 5:00 the next day, the morning cooling period from 6:00 to 11:00, the midday high load period from 12:00 to 17:00, and the night transition period from 18:00 to 22:00.The control range of cold storage temperature varies with the operating time period. The inlet water temperature is maintained at 5 to 7 degrees Celsius during the late night cold storage period, and is allowed to rise to 8 to 10 degrees Celsius during the cooling period; the outlet water temperature is controlled at 12 to 14 degrees Celsius during the late night cold storage period, and can rise to 15 to 17 degrees Celsius during the cooling period. The main frequency component analysis shows that the main period of temperature fluctuation is 6 hours. The adjustment range of cold storage power is related to the capacity of the device. For a cold storage device with a rated capacity of 1,000 kW, when the inlet and outlet water temperature difference is 8 degrees Celsius under standard working conditions, the lower limit of the cold storage power is 400 kW and the upper limit is 800 kW. The calculation of thermal enthalpy difference shows that about 0.15 cubic meters of chilled water are required for each kilowatt-hour of cold storage. The kernel density estimation method is used to analyze the cold storage data, and the bandwidth parameter is selected as 50 kilowatt-hour. The density distribution curve obtained shows a bimodal feature. The main peak corresponds to the daily cooling capacity, which is located at 75% of the rated capacity; the secondary peak corresponds to the cooling capacity during the high-load period, which is located at 90% of the rated capacity. This distribution feature provides an important basis for determining the boundary of the cooling storage capacity.

[0027] Step S104, according to the dynamic adjustment parameters of cold storage, combined with the performance parameters and operating constraints of the cold storage system, through the energy efficiency analysis method, analyze the impact mechanism of the cold storage adjustment plan on the operating efficiency of the cold storage system, the impact mechanism includes the impact of the adjustment plan on the cold storage capacity, cold release capacity and energy consumption of the cold storage system, and obtain the change characteristics of the system efficiency, including the comprehensive energy efficiency of the system, the operating cost under peak and valley electricity prices, and the carbon emission intensity.

[0028] According to the cold storage capacity parameters, the cold release capacity parameters and the power consumption parameters, the adaptive weighted regression method is used for coupling calculation to obtain the refrigeration coefficient curve; the refrigeration coefficient curve is subjected to piecewise linear fitting, the slope and intercept of each segment are extracted, and a refrigeration coefficient prediction function is constructed to obtain the energy efficiency prediction value; the random forest algorithm is used to process the refrigeration coefficient curve and the peak and valley electricity price data to obtain the power consumption cost value; according to the power consumption cost value and the carbon emission factor data of the power grid, the carbon emissions per unit time are calculated in combination with the power consumption curve of the unit to form a carbon emission curve.

[0029] Exemplarily, according to the operating parameter data sequence of the cold storage device, an adaptive weighted regression method is used to couple the cold storage capacity, the cold release capacity, and the power consumption, and a thermodynamic process balance equation is constructed to obtain a first refrigeration coefficient curve. A piecewise linear fitting is performed on the first refrigeration coefficient curve, and the slope and intercept of each segment are extracted to construct a refrigeration coefficient prediction function to obtain a first energy efficiency prediction value. The real-time peak and valley electricity price values ​​and the time-of-use electricity price period data are sampled, and an operation cost prediction function is constructed using a random forest algorithm. The first power consumption cost value is calculated by inputting the cold storage time, cold storage power, and cold storage temperature parameters. According to the first power consumption cost value and the grid carbon emission factor, combined with the unit power consumption curve, the carbon emissions per unit time are calculated to obtain a first carbon emission curve. The inlet and outlet water temperature parameters and flow parameters of the cold storage device are used, combined with the water pump power curve, to construct a thermal enthalpy calculation function of the cold storage process, and the ratio of the cold storage capacity to the cold release capacity per unit time is calculated. According to the ratio of the cold storage capacity to the cold release capacity per unit time, a cooling efficiency calculation function is constructed to obtain a second energy efficiency prediction value. The first energy efficiency prediction value, the second energy efficiency prediction value, the first power consumption cost value, and the first carbon emission curve are normalized, and the index weight coefficient is calculated by the hierarchical analysis method. According to the normalized index values ​​and weight coefficients, the comprehensive energy efficiency evaluation value is obtained by weighted summation. The refrigeration coefficient of the cold storage device represents the ratio of cooling capacity to power consumption. Under standard working conditions, the cold storage capacity is 1000 kWh, the power consumption is 200 kW, the cooling capacity is 800 kWh, and the refrigeration coefficient of the device is 4. Through adaptive weighted regression processing of operating data, it is found that the refrigeration coefficient presents a segmented change feature. In the cold storage temperature range of 5 to 7 degrees Celsius, the refrigeration coefficient decreases with increasing temperature, with a slope of -0.5 per degree Celsius. The time-of-use electricity price structure has a significant impact on the operating cost. The peak-to-valley electricity price ratio is usually 3:1, the peak period electricity price is 1.2 yuan per kilowatt-hour, and the valley period electricity price is 0.4 yuan per kilowatt-hour. The prediction of the random forest algorithm shows that at the same cold storage temperature, if the cold storage period is adjusted from the peak period to the valley period, the operating cost will be reduced by 65%. The carbon emission factor of the power grid varies with the period. The carbon emission factor of the night valley period is 0.8 kg per kilowatt-hour, and the carbon emission factor of the daytime peak period is 0.6 kg per kilowatt-hour. For a cold storage device with a rated power of 300 kilowatts, the carbon emissions per hour during the peak period are 180 kg, and the carbon emissions per hour during the valley period are 240 kg. The change in enthalpy of the cold storage process is closely related to the water temperature. When the inlet water temperature is 5 degrees Celsius and the outlet water temperature is 12 degrees Celsius, the enthalpy difference per unit mass of water is 29.3 kilojoules per kilogram. Under the condition of a circulating water flow of 50 cubic meters per hour, the cold storage rate is 420 kilowatts and the cooling rate is 380 kilowatts. The pump power curve shows that for every increase of 10 cubic meters per hour in flow, the power consumption increases by 5 kilowatts. The analytic hierarchy process assigns weights to different indicators, with the weight of the cooling coefficient being 0.4, the weight of the operating cost being 0.35, and the weight of the carbon emission being 0.25.When the cooling coefficient increases from 3.5 to 4.0, the operating cost decreases by 20%, and the carbon emissions increase by 10%, the comprehensive energy efficiency evaluation value increases by 15%. This multi-index evaluation system reflects the comprehensive performance of the cold storage device in terms of energy conservation, emission reduction and economy. Through the dynamic adjustment of operating parameters, the cold storage device exhibits different performance characteristics. Under summer conditions, when the cold storage temperature drops from 7 degrees Celsius to 5 degrees Celsius, the cooling coefficient decreases by 0.8, but the cold storage capacity increases by 25%. In the transition season, the cold storage temperature is maintained at 6 degrees Celsius, and the cold storage time is adjusted from 6 hours to 8 hours, and the comprehensive energy efficiency evaluation value increases by 12%.

[0030] Step S105, judging whether the operating state of the cold storage system meets the preset threshold conditions through the changing characteristics of the system efficiency; if the improvement of the system comprehensive energy efficiency exceeds the preset threshold, the reduction of the operating cost exceeds the preset threshold, and the reduction of the carbon emission intensity exceeds the preset threshold, it is judged that the threshold conditions are met and the current strategy adjustment plan is output.

[0031] The indicator change rate within each sampling period is calculated according to the comprehensive energy efficiency value, operating cost value and carbon emission intensity value to obtain the indicator change rate sequence; the indicator change rate sequence is segmented by a sliding time window, and the change amplitude of adjacent sampling moments is calculated for the data in each time window to obtain the efficiency change sequence; the gradient descent method is applied to the efficiency change sequence, and the mean square error is used as the optimization objective function to calculate the time series change trend of the efficiency improvement amplitude, cost reduction amplitude and emission reduction amplitude; a three-dimensional feature vector is constructed according to the time series change trend, and the distance value from the feature vector to the preset threshold plane is calculated by the radial basis kernel function of the support vector machine to obtain a time interval sequence that meets the threshold condition.

[0032] Exemplarily, according to the comprehensive energy efficiency value, operating cost value and carbon emission intensity value in the real-time operation data sequence, the index change rate in each sampling period is calculated to obtain the first index change rate sequence. The first index change rate sequence is segmented using a sliding time window with a fixed length of 24 hours, and the change amplitude of two adjacent sampling moments is calculated for the data in each time window to obtain the first efficiency change sequence. The gradient descent method is applied to the first efficiency change sequence, and the mean square error is used as the optimization objective function to calculate the time series change trend of the efficiency improvement amplitude, cost reduction amplitude and emission reduction amplitude. A three-dimensional feature vector is constructed according to the time series change trend, and the radial basis kernel function of the support vector machine is used to calculate the distance value from the feature vector to the preset threshold plane. By screening the forward projection distance of the preset threshold plane, a time interval sequence that meets the threshold condition is obtained. In the time interval that meets the threshold condition, the sampled values ​​of the cold storage duration parameter, the cold storage power parameter and the cold storage temperature parameter are extracted. The extracted parameter sample values ​​are detected for abnormal values, and the abnormal data points that exceed the preset range are eliminated to obtain the parameter optimization sequence. The kernel density estimation method is used to fit the probability distribution of the parameter optimization sequence, and the parameter combination corresponding to the maximum probability density is extracted. The evaluation of the operation index of the cold storage system involves multiple dimensions. When the sampling frequency is 5 minutes, the comprehensive energy efficiency value fluctuates between 3.5 and 4.5, the operating cost varies between 0.8 and 1.2 yuan per kilowatt-hour, and the carbon emission intensity is distributed between 0.6 and 0.9 kg per kilowatt-hour. By calculating the difference between adjacent sampling points, the energy efficiency improvement is 0.02 to 0.05 per sampling cycle. The 24-hour sliding time window reflects the daily cycle characteristics of the index change. The operating cost change rate during the daytime is positive and negative during the nighttime, reflecting the impact of peak and valley electricity prices. The carbon emission intensity decreases significantly during the photovoltaic power generation period, with a change rate of -0.05 kg per kilowatt-hour per hour. The comprehensive energy efficiency value increases fastest in the early morning when the temperature is lower, with a change rate of 0.08 per hour. The distance from the three-dimensional feature vector to the preset threshold plane reflects the degree of optimization of the operating state. The threshold plane is determined by three critical values: energy efficiency improvement of 0.3, cost reduction of 20%, and carbon emission reduction of 15%. When the forward projection distance of the eigenvector is greater than 0.1, it indicates that the direction of adjustment of the operating parameters is reasonable. Actual operation data show that the time interval that meets the threshold conditions is mainly distributed from 23:00 at night to 5:00 the next day. The definition of abnormal values ​​of cold storage parameters is based on physical constraints. The cold storage time is between 2 and 8 hours, the cold storage power is between 300 and 900 kilowatts, and the cold storage temperature is between 4 and 8 degrees Celsius. Data points outside this range are marked as outliers. Statistics show that outliers account for 3% of the total samples, mainly appearing in the start and stop process of the equipment. The probability distribution of the parameter optimization sequence presents multi-peak characteristics, and the kernel density estimation uses the Gaussian kernel function with a bandwidth of 0.5. The main peak of the distribution curve corresponds to the parameter combination of 6 hours of cold storage time, 600 kilowatts of cold storage power, and 6 degrees Celsius of cold storage temperature.The parameter combination corresponding to the sub-peak reflects the difference in seasonal operating conditions. The summer operating conditions correspond to a cold storage time of 4 hours, a cold storage power of 800 kilowatts, and a cold storage temperature of 5 degrees Celsius. The optimal parameter combination shows good adaptability in different operating scenarios. Under standard conditions, this parameter combination improves the overall energy efficiency by 0.4, reduces operating costs by 25%, and reduces carbon emissions by 18%. Under partial load conditions, similar optimization effects were maintained by adjusting the cold storage time to 5 hours. In the transition season, the cold storage temperature was appropriately increased to 7 degrees Celsius, and the improvement trend of various indicators was still maintained.

[0033] Step S106, applying the strategy adjustment scheme that meets the threshold conditions to the cold storage system, monitoring the changing trend of the system efficiency in real time, and generating a cold storage system efficiency evaluation report in combination with the big data analysis method. The evaluation report includes the correlation between the shielding area, power generation efficiency, cold storage and system efficiency.

[0034] A time series predictor is constructed based on the monitored data using a long short-term memory network. The time series predictor extracts time series features for each indicator to obtain an indicator sequence. A sliding correlation window is constructed for the indicator sequence. The sliding correlation window calculates the Pearson correlation coefficient between indicators to obtain a correlation coefficient matrix. An indicator mapping function is constructed based on the correlation coefficient matrix using a random forest regression algorithm. The indicator mapping function calculates the one-way influence intensity between indicators to obtain an influence intensity matrix. An indicator transfer network is constructed based on the influence intensity matrix using a feature vector sequence. The indicator transfer network calculates the in-degree value and out-degree value of each node to obtain an indicator influence intensity ranking. A hierarchical indicator association diagram is generated based on the indicator influence intensity ranking, the transmission path from the blocked area to the comprehensive efficiency is extracted, and a performance evaluation report including indicator association relationships, influence levels, and response characteristics is generated.

[0035] Exemplarily, according to the online monitoring data, the shielding area value, power generation efficiency value, cold storage value and comprehensive efficiency value are collected, the sampling period is set to 5 minutes, and the long short-term memory network is used to construct a time series predictor to extract the time series characteristics of each indicator to obtain the first indicator sequence. The first indicator sequence is subjected to time series correlation analysis, and a sliding correlation window is constructed to calculate the Pearson correlation coefficient between the indicators to obtain the first correlation coefficient matrix. According to the first correlation coefficient matrix, significant correlation items are extracted, and the indicator mapping function is constructed by the random forest regression algorithm to calculate the one-way influence intensity between the indicators to obtain the first influence intensity matrix. The first influence intensity matrix is ​​subjected to singular value decomposition, the main eigenvectors are extracted, and the eigenvalue contribution rate is calculated to obtain the first eigenvector sequence. The indicator transfer network is constructed using the first eigenvector sequence, and the in-degree value and out-degree value of each node in the network are calculated to obtain the indicator influence intensity ranking. According to the indicator influence intensity ranking, a hierarchical indicator association diagram is generated, and the association intensity value is marked to obtain the first association structure diagram. The first association structure diagram is subjected to path analysis to extract the key transfer path from the shielding area to the comprehensive efficiency, and a transfer chain diagram is generated. The transfer chain diagram is used to quantitatively evaluate the current operation strategy and generate a performance evaluation report containing indicator correlation, impact degree, and response characteristics. The operating indicators of the photovoltaic cold storage system show obvious time series correlation. When the sampling period is 5 minutes, the power generation efficiency decreases by 15% when the shaded area increases from 5% to 10%, resulting in a 12% decrease in cold storage power, and ultimately an 8% decrease in overall efficiency. The prediction results of the long short-term memory network show that the change in shaded area leads the change in power generation efficiency by about 10 minutes, and the change in power generation efficiency leads the change in cold storage by about 15 minutes. When the sliding correlation window is set to 1 hour, the Pearson correlation coefficients between the indicators show significant characteristics. The correlation coefficient between shaded area and power generation efficiency is -0.85, showing a strong negative correlation; the correlation coefficient between power generation efficiency and cold storage is 0.78, showing a strong positive correlation; the correlation coefficient between cold storage and overall efficiency is 0.82, also showing a strong positive correlation. Random forest regression analysis reveals the quantitative influence relationship between indicators. For every 1 percentage point increase in the shading area, the power generation efficiency decreases by 1.5 percentage points; for every 1 percentage point decrease in the power generation efficiency, the cold storage decreases by 0.8 percentage points; for every 1 percentage point decrease in the cold storage, the overall efficiency decreases by 0.6 percentage points. This chain transfer feature becomes clearer after singular value decomposition. The eigenvalue analysis shows that the contribution rate of the first principal component reaches 75%, and the corresponding eigenvector reflects the dominant position of the shading effect. In the indicator transfer network, the out-degree value of the shading area node is the largest, which is 3; the out-degree value and in-degree value of the power generation efficiency node are both 2; the in-degree values ​​of the cold storage node and the overall efficiency node are 2 and 3 respectively. The hierarchical indicator association diagram clearly shows the performance transfer path. The main path is that the increase in the shading area leads to weakened light, the weakened light causes the power generation power to decrease, the power decrease causes the cold storage capacity to decrease, and finally affects the overall efficiency.Secondary paths include the shaded area affecting power generation efficiency through temperature effects, and power generation efficiency affecting cold storage through peak-valley regulation. Quantitative evaluation results show that under standard operating conditions, by optimizing the cold storage strategy, when the shaded area increases by 10%, the decrease in power generation efficiency is reduced from 15% to 12%, the decrease in cold storage is reduced from 12% to 9%, and the decrease in comprehensive efficiency is controlled from 8% to within 6%. This improvement effect is particularly obvious when the continuous shade time exceeds 2 hours.

[0036] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. An efficiency evaluation method for photovoltaic power generation cold storage system based on big data analysis, characterized in that: The method comprises: Obtain the shadow distribution data on the surface of the photovoltaic module, extract the change characteristics of the shielding area through image processing methods, and judge the dynamic change trend of the shielding area by combining the time series analysis method; according to the dynamic change trend of the shielding area, use the regression model to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic module, and establish a mapping relationship between the power generation efficiency and the shielding area; according to the fluctuation characteristics of the power generation efficiency, analyze the historical operation data and meteorological data of the photovoltaic power generation system, extract the change law of the cold storage demand of the cold storage system, and determine the adjustment content of the cold storage strategy. The adjustment content includes dynamic adjustment parameters, and the dynamic adjustment parameters include adjustment of cold storage time, cold storage power and cold storage temperature parameters; according to the dynamic adjustment parameters of cold storage, combined with the performance parameters and operation constraints of the cold storage system, analyze the impact of the cold storage adjustment plan on the operation of the cold storage system through energy efficiency analysis methods. The influencing mechanism of efficiency includes the influence of the adjustment plan on the cold storage capacity, cold release capacity and energy consumption of the cold storage system, and the change characteristics of the system efficiency are obtained. The change characteristics include the system comprehensive energy efficiency, the operating cost under the peak and valley electricity prices, and the carbon emission intensity. Through the change characteristics of the system efficiency, it is judged whether the operating status of the cold storage system meets the preset threshold conditions. If the improvement of the system comprehensive energy efficiency exceeds the preset threshold, the reduction of the operating cost exceeds the preset threshold, and the reduction of the carbon emission intensity exceeds the preset threshold, it is judged that the threshold conditions are met, and the current strategy adjustment plan is output. The strategy adjustment plan that meets the threshold conditions is applied to the cold storage system, and the changing trend of the system efficiency is monitored in real time. Combined with the big data analysis method, an efficiency evaluation report of the cold storage system is generated. The evaluation report includes the correlation between the shielding area, power generation efficiency, cold storage and system efficiency.

2. The method according to claim 1, characterized in that The method of obtaining shadow distribution data on the surface of the photovoltaic module, extracting the change characteristics of the shielding area by an image processing method, and judging the dynamic change trend of the shielding area by combining a time series analysis method includes: Using an image acquisition device to acquire a high-resolution image of the surface of the photovoltaic module, dividing the shadow area according to a comparison result between the grayscale value of a pixel point of the high-resolution image and a preset reference grayscale value, and obtaining a first shadow area value; Performing edge detection on the shadow area to obtain an edge contour line, using the least squares method to fit the edge contour line to obtain an edge contour curve equation, establishing an autoregressive moving average model based on the edge contour curve equation, and obtaining a motion trajectory equation of the shadow area; The temperature distribution matrix is ​​constructed by collecting temperature data through a pre-deployed temperature sensor array, and isotherms are extracted for the temperature distribution matrix. If the temperature value in the isotherm closed curve is lower than the temperature value in the surrounding area, it is determined that there is a shadow in the area, and a second shadow area value is obtained; The weighted average method is used to fuse the first shadow area value and the second shadow area value, and the fused shadow area value sequence is subjected to discrete Fourier transform. The frequency components with amplitudes greater than the normalized threshold are extracted according to the spectrum diagram of the discrete Fourier transform to obtain a shadow area change periodic sequence.

3. The method according to claim 1, characterized in that The method uses a regression model to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic module according to the dynamic change trend of the shielding area, and establishes a mapping relationship between the power generation efficiency and the shielding area, including: Extracting the fluctuation component in the power generation efficiency sequence through a sliding time window to obtain a first power generation efficiency curve; According to the first power generation efficiency curve, a nonlinear mapping function is constructed by using a support vector regression method to obtain a first prediction function; For the first prediction function, calculating a temperature correction coefficient according to a temperature coefficient to obtain a second prediction function; The second prediction function is used to process the real-time shielding area data to obtain a second power generation efficiency curve, and a high-frequency component and a low-frequency component in the second power generation efficiency curve are extracted by wavelet transform to obtain a third power generation efficiency curve; The third power generation efficiency curve is subjected to Fourier transformation, the spectrum feature vector is extracted, and the ambient temperature and humidity data are integrated using the Bayesian regression method to predict the future power generation efficiency change trend curve; it also includes: according to the dynamic change trend of the shielding area, a regression model with the shielding area as the independent variable and the power generation efficiency as the dependent variable is established, the fluctuation characteristics of the power generation efficiency of the photovoltaic module with the change of the shielding area are predicted, and the quantitative mapping relationship between the power generation efficiency and the shielding area is fitted, including the change range of the shielding area, the fluctuation amplitude of the power generation efficiency and the peak efficiency point, so as to predict the power generation efficiency change curve under different shielding scenarios.

4. The method according to claim 3, characterized in that According to the dynamic change trend of the shielding area, a regression model with the shielding area as the independent variable and the power generation efficiency as the dependent variable is established to predict the fluctuation characteristics of the power generation efficiency of the photovoltaic module with the change of the shielding area, and fit the quantitative mapping relationship between the power generation efficiency and the shielding area, including the change range of the shielding area, the fluctuation amplitude of the power generation efficiency and the peak efficiency point, so as to predict the power generation efficiency change curve under different shielding scenarios, including: The light intensity data is collected according to the light intensity sensor array, a light intensity distribution matrix is ​​constructed for the light intensity data, and a threshold segmentation method is used to process the light intensity distribution matrix to obtain a binary image of the occluded area; the pixel ratios of the occluded area in the binary image of the occluded area are counted to obtain a first numerical sequence of occluded area, and the first numerical sequence of occluded area is subjected to Fourier transformation to extract spectral features, and a trigonometric function superposition fitting is used to obtain a dynamic change function of the occluded area; the voltage and current sampling data output by the photovoltaic module are obtained, the power generation value is calculated according to the voltage and current sampling data, and a fixed time window is used to perform sliding average processing on the power generation value to obtain a first power generation efficiency curve; the sampling value of the dynamic change function of the occluded area is used as an independent variable, and the corresponding time value of the first power generation efficiency curve is used as a dependent variable, and a first prediction function is obtained by training with a support vector regression method.

5. The method according to claim 1, characterized in that The method aims at analyzing the historical operation data and meteorological data of the photovoltaic power generation system in view of the fluctuation characteristics of power generation efficiency, extracting the changing law of the cold storage demand of the cold storage system, and determining the adjustment content of the cold storage strategy, wherein the adjustment content includes dynamic adjustment parameters, and the dynamic adjustment parameters include adjusting the cold storage time, cold storage power and cold storage temperature parameters, including: According to the photovoltaic power generation power sampling data, the sampling data is processed by a fast Fourier transform method to obtain the periodic characteristics and amplitude characteristics of the power generation power fluctuation; The periodic characteristics and amplitude characteristics are processed in segments, and a feature extraction method is used to obtain meteorological feature vectors of light intensity, ambient temperature and relative humidity from the segmented data to obtain a feature sequence; According to the characteristic sequence, a support vector machine clustering method is used to process the characteristic sequence to obtain a cold storage load prediction curve; According to the cold storage load prediction curve, a linear programming solver is used to calculate the start and end points of the cold storage time period to obtain a cold storage duration curve; The cold storage time curve is segmented and integrated to calculate the cold storage power demand in each cold storage time period; it also includes: mining the changing patterns of cold storage system cold demand over time and weather in historical operation and meteorological data, identifying key parameters affecting cold storage demand, extracting intraday fluctuations and seasonal trends in load demand, and determining the optimization space of cold storage strategy based on changes in cold demand, adjusting the action space of cold storage operation parameters, including adjustable time period range, upper and lower power limits, and temperature setting allowable range, to form the boundary of dynamic optimization of cold storage strategy parameters.

6. The method according to claim 5, characterized in that The method mines the changing rules of the cold demand of the cold storage system over time and weather in the historical operation and meteorological data, identifies the key parameters affecting the cold storage demand, extracts the intraday fluctuation and seasonal change trend of the load demand, determines the optimization space of the cold storage strategy based on the change of the cold demand, adjusts the action space of the cold storage operation parameters, including the adjustable time range, the upper and lower limits of power and the allowable range of temperature setting, and forms the boundary of the dynamic optimization of the cold storage strategy parameters, including: According to the cooling load sequence and the meteorological parameter sequence, a long short-term memory network is used to extract the intraday fluctuation law and seasonal change characteristics of the cooling load to obtain a first load fluctuation curve; the first load fluctuation curve is subjected to wavelet transform processing, and the second load fluctuation curve is obtained by extracting high-frequency fluctuation components and low-frequency change trends; the second load fluctuation curve and the meteorological parameter sequence are processed by a decision tree regressor, and the load influence parameter sequence is obtained by calculating the influence of temperature, humidity, and light intensity on load changes; the cold storage device operation data is classified according to the load influence parameter sequence, and the boundary points of the cold storage period are identified by the hierarchical clustering method of Euclidean distance to obtain the cold storage operation segmentation result; the inlet and outlet water temperature data sequence in each cold storage operation segment is subjected to Fourier transform, and the temperature fluctuation range corresponding to the main frequency component is extracted to obtain the upper and lower limits of the cold storage temperature; The thermal enthalpy difference is calculated according to the upper and lower limits of the cold storage temperature, and the cold storage power adjustment range is obtained in combination with the rated capacity of the cold storage device.

7. The method according to claim 1, characterized in that According to the dynamic adjustment parameters of cold storage, combined with the performance parameters and operation constraints of the cold storage system, the influence mechanism of the cold storage adjustment scheme on the operation efficiency of the cold storage system is analyzed through the energy efficiency analysis method. The influence mechanism includes the influence of the adjustment scheme on the cold storage capacity, cold discharge capacity and energy consumption of the cold storage system, and the change characteristics of the system efficiency are obtained. The change characteristics include the comprehensive energy efficiency of the system, the operation cost under the peak and valley electricity prices, and the carbon emission intensity, including: According to the parameters of cold storage capacity, cold release capacity and power consumption, the adaptive weighted regression method is used for coupling calculation to obtain the refrigeration coefficient curve; Perform piecewise linear fitting on the cooling coefficient curve, extract the slope and intercept of each segment, construct the cooling coefficient prediction function, and obtain the energy efficiency prediction value; The random forest algorithm is used to process the cooling coefficient curve and the peak and valley electricity price data to obtain the electricity consumption cost value; The carbon emissions per unit time are calculated based on the electricity consumption cost value and the grid carbon emission factor data in combination with the unit electricity consumption power curve to form a carbon emission curve.

8. The method according to claim 1, characterized in that The change characteristics of the system efficiency are used to determine whether the operating state of the cold storage system meets the preset threshold conditions. If the system comprehensive energy efficiency improvement exceeds the preset threshold, the operating cost reduction exceeds the preset threshold, and the carbon emission intensity reduction exceeds the preset threshold, it is determined that the threshold conditions are met, and the current strategy adjustment plan is output, including: The index change rate in each sampling period is calculated based on the comprehensive energy efficiency value, operating cost value and carbon emission intensity value to obtain the index change rate sequence; The index change rate sequence is segmented using a sliding time window, and the change amplitude of adjacent sampling moments is calculated for the data in each time window to obtain an efficiency change sequence; Applying the gradient descent method to the efficiency change sequence, and using the mean square error as the optimization objective function to calculate the time series change trend of the efficiency improvement range, the cost reduction range, and the emission reduction range; A three-dimensional feature vector is constructed according to the time series variation trend, and the distance value from the feature vector to a preset threshold plane is calculated by the radial basis kernel function of the support vector machine to obtain a time interval sequence that meets the threshold condition.

9. The method according to claim 1, characterized in that: The strategy adjustment scheme that meets the threshold conditions is applied to the cold storage system, and the changing trend of the system efficiency is monitored in real time. A cold storage system efficiency evaluation report is generated in combination with a big data analysis method. The evaluation report includes the correlation between the shielding area, power generation efficiency, cold storage and system efficiency, including: A time series predictor is constructed using a long short-term memory network according to the monitored data, and the time series predictor extracts time series features of each indicator to obtain an indicator sequence; Constructing a sliding correlation window for the indicator sequence, wherein the sliding correlation window calculates the Pearson correlation coefficient between the indicators to obtain a correlation coefficient matrix; An indicator mapping function is constructed based on the correlation coefficient matrix using a random forest regression algorithm, and the indicator mapping function calculates the one-way influence intensity between the indicators to obtain an influence intensity matrix; An indicator transfer network is constructed using a characteristic vector sequence for the influence intensity matrix, and the indicator transfer network calculates the in-degree value and out-degree value of each node to obtain an indicator influence intensity ranking; A hierarchical indicator correlation diagram is generated according to the ranking of indicator impact intensity, the transmission path from the shielding area to the comprehensive efficiency is extracted, and a performance evaluation report containing indicator correlation, impact degree and response characteristics is generated.

Images