Energy efficiency evaluation optimization method and system for photovoltaic power generation cold storage system

Abstract

The invention provides an energy efficiency conversion evaluation method for a photovoltaic power generation cold storage system, and the method comprises the steps: obtaining the real-time power data and weather condition data of a photovoltaic power station, and calculating a power fluctuation ratio sequence; extracting characteristic values from the power fluctuation ratio sequence, and determining an influence coefficient sequence of fluctuation on ice-making energy input; obtaining temperature data and water volume data in the ice storage tank, updating a real-time energy input model through the influence coefficient sequence, and calculating dynamic changes of a temperature sequence and a water volume sequence; according to dynamic changes of the temperature sequence and the water quantity sequence, a cooling efficiency sequence is constructed; building load demand data and environment temperature data are acquired, and a load demand change rate sequence is calculated; a dynamic coupling relation sequence is constructed through the cooling efficiency sequence and the load demand change rate sequence, and an energy transfer delay sequence is determined; and extracting full-period features from the dynamic coupling relation sequence and the energy transfer delay sequence, and generating an energy efficiency prediction sequence.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to an energy efficiency evaluation and optimization method and system for a photovoltaic power generation and ice storage cooling system. Background Art

[0002] In the energy efficiency evaluation and optimization of a photovoltaic power generation and ice storage cooling system, existing methods generally focus on static models and fail to fully consider the complex interaction between the volatility of photovoltaic power output and the dynamic operation of the ice storage cooling system. The power generation of photovoltaic power is driven by external factors such as weather conditions and sunlight intensity, showing significant uncertainty and time-variation. This volatility directly affects the energy input during the ice-making and ice storage process, making it difficult to stabilize the real-time state of the water temperature and water volume in the ice storage tank. The dynamic changes in water temperature and water volume further affect the efficiency of the ice melting and cooling process. The cooling efficiency not only depends on the current state of the ice storage tank but is also closely coupled with the real-time demand of the external cooling load. The load demand may fluctuate violently within a short period due to factors such as building usage patterns and environmental temperature changes, resulting in time lags or mismatches in the energy transfer between ice-making and cooling. Traditional static models usually assume that parameters such as input power, water temperature, and water volume are fixed values, ignoring the non-linear changes of these variables during actual operation and their feedback effects. This assumption leads to evaluation results that cannot reflect the true energy efficiency performance of the system at different time scales, especially in extreme weather or load mutation scenarios, where the deviation is particularly significant. Therefore, it is necessary to accurately quantify the instantaneous impact of photovoltaic power volatility on the ice-making and ice storage energy input, consider the dynamic coupling relationship between the water temperature and water volume in the ice storage tank and the cooling load demand, and on this basis, overcome the limitations of traditional static models in multi-variable real-time interaction scenarios, and construct an evaluation framework that can capture the energy efficiency characteristics of the system throughout its life cycle. Summary of the Invention

[0003] The present invention provides an energy efficiency evaluation and optimization method for a photovoltaic power generation and ice storage cooling system, mainly including: Obtain the real-time power data and weather condition data of a photovoltaic power station, and calculate the power volatility sequence; extract eigenvalue sequences from the power volatility sequence to determine the influence coefficient sequence of fluctuations on ice-making energy input; obtain the temperature data and water volume data in the ice storage tank, update the real-time energy input model through the influence coefficient sequence, and calculate the dynamic changes in the temperature sequence and water volume sequence; construct a cooling efficiency sequence based on the dynamic changes in the temperature sequence and water volume sequence; obtain the building load demand data and environmental temperature data, and calculate the load demand change rate sequence; construct a dynamic coupling relationship sequence through the cooling efficiency sequence and the load demand change rate sequence, and determine the energy transfer delay sequence; extract full-cycle characteristics from the dynamic coupling relationship sequence and the energy transfer delay sequence to generate an energy efficiency prediction sequence.

[0004] Further, the steps of obtaining real-time power data and weather condition data of the photovoltaic power station and calculating the power volatility sequence include: collecting photovoltaic output power and sunshine intensity through sensors, calculating the power volatility using the time series analysis method, and if the power at the previous time point is zero and the current power is non-zero, setting the power volatility to a preset value; processing the power volatility and sunshine intensity through a sliding window method, calculating the average power volatility and average sunshine intensity within a continuous time period, and analyzing the correlation between the average power volatility and average sunshine intensity using the Pearson correlation coefficient to obtain the power volatility quantization result.

[0005] Further, the steps of extracting eigenvalue from the power volatility sequence and determining the influence coefficient sequence of the fluctuation on the ice-making energy input include: processing the power volatility sequence through a sliding window method to extract the smooth fluctuation feature sequence; calculating the energy input per unit time through integration, and analyzing the instantaneous mapping relationship between the smooth fluctuation feature sequence and the energy input using a convolutional neural network to obtain the mapping feature matrix; calculating the influence coefficient sequence according to the mapping feature matrix, and if the influence coefficient exceeds the preset threshold, processing the relationship between the smooth fluctuation feature sequence and the energy input through the random forest algorithm to adjust the influence coefficient sequence to obtain the dynamic influence sequence of the fluctuation on the energy input.

[0006] Further, the steps of updating the real-time energy input model through the influence coefficient sequence and calculating the dynamic changes of the temperature sequence and water volume sequence include: calculating and updating the energy sequence through the influence coefficient sequence and the energy input, generating the temperature sequence and water volume sequence using a difference equation; processing the temperature sequence and water volume sequence through a sliding window method to obtain the smooth temperature sequence and smooth water volume sequence; analyzing the mapping relationship between the smooth temperature sequence and the updated energy sequence using linear regression to obtain the temperature-energy correlation coefficient matrix; generating a comprehensive dynamic influence matrix by combining the temperature-energy correlation coefficient matrix and the water volume-energy correlation coefficient matrix through a matrix fusion method, and adjusting the temperature sequence and water volume sequence according to the comprehensive dynamic influence matrix.

[0007] Further, constructing a cooling efficiency sequence according to the dynamic changes of the temperature sequence and the water volume sequence includes: calculating an initial cooling efficiency sequence through a preset efficiency weight coefficient, inputting the initial cooling efficiency sequence and historical data into a support vector machine model to generate an efficiency prediction sequence; the efficiency weight coefficient is determined by multiple linear regression of historical cooling data, where the temperature weight γ ranges from 0.32 to 0.45, and the water volume weight ranges from 0.55 to 0.68; processing the efficiency prediction sequence by a sliding window method to obtain a smoothed efficiency sequence; analyzing the linear relationship between the smoothed efficiency sequence and the temperature sequence by the least squares method to obtain a temperature efficiency correlation coefficient matrix; generating a comprehensive efficiency influence matrix by matrix-weighted fusion of the temperature efficiency correlation coefficient matrix and the water volume efficiency correlation coefficient matrix, and adjusting the smoothed efficiency sequence according to the comprehensive efficiency influence matrix.

[0008] Further, obtaining building load demand data and ambient temperature data and calculating a load demand change rate sequence includes: decomposing the load demand data through a long short-term memory network to obtain a trend component and a periodic component, and calculating a load demand change rate sequence; analyzing the correlation between the load demand change rate sequence and the ambient temperature data by the Pearson correlation coefficient to obtain a correlation coefficient sequence; processing the correlation coefficient sequence by a sliding window method to obtain a smoothed correlation coefficient sequence; fitting the non-linear relationship between the smoothed correlation coefficient sequence and the ambient temperature data by a random forest model to generate a fitting relationship function, and adjusting the load demand change rate sequence according to the fitting relationship function.

[0009] Further, constructing a dynamic coupling relationship sequence through the cooling efficiency sequence and the load demand change rate sequence and determining an energy transfer delay sequence includes: calculating a dynamic coupling relationship sequence through a preset coupling weight, and if the dynamic coupling relationship sequence exceeds a preset threshold, calculating a time-delay effect feature by a sliding window method to generate an energy transfer delay sequence; processing the dynamic coupling relationship sequence and the energy transfer delay sequence by a support vector machine algorithm to obtain a mapping relationship function; analyzing the mapping relationship function and the cooling efficiency sequence by a long short-term memory network to predict the dynamic coupling relationship sequence in the next time period, and adjusting the dynamic coupling relationship sequence according to the prediction result.

[0010] Further, extracting full-cycle features from the dynamic coupling relationship sequence and the energy transfer delay sequence to generate an energy efficiency prediction sequence includes: extracting the periodic features of the dynamic coupling relationship sequence and the energy transfer delay sequence through a preprocessing method, and using a random forest algorithm to perform multivariate regression on the temperature sequence, water volume sequence, cooling efficiency sequence, and load demand change rate sequence to generate a preliminary energy efficiency prediction sequence; calculating the fluctuation feature sequence of the preliminary energy efficiency prediction sequence through a sliding window method. If the fluctuation feature sequence exceeds a preset threshold, then use a support vector machine algorithm to process the fluctuation feature sequence and the energy transfer delay sequence to obtain a mapping feature sequence, and adjust the energy efficiency prediction sequence according to the mapping feature sequence.

[0011] The technical solution provided by the embodiment of the present invention may include the following beneficial effects: The present invention discloses an energy efficiency evaluation and optimization method for a photovoltaic power generation and ice storage cooling system. The method collects photovoltaic power and weather data in real time, analyzes the impact of power fluctuations on the ice-making energy input, and combines the ice storage tank temperature and water volume changes to construct a dynamic ice melting and cooling efficiency model. At the same time, the system uses deep learning methods to predict building load demand, establishes a dynamic coupling relationship between cooling efficiency and load demand, and considers the time delay effect. The present invention also uses multivariate regression analysis and adaptive filtering technology to achieve energy efficiency prediction and model parameter optimization at different time scales. Description of the Drawings

[0012] Figure 1 It is a flowchart of an energy efficiency evaluation and optimization method for a photovoltaic power generation and ice storage cooling system of the present invention. Detailed Embodiments

[0013] To further understand the content of the present invention, the present invention will be described in detail in combination with the drawings and embodiments. The following will further elaborate on the present application with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention and are not intended to limit the invention. Additionally, it should be noted that for the sake of description, only parts related to the invention are shown in the drawings.

[0014] As Figure 1 , an energy efficiency evaluation and optimization method for a photovoltaic power generation and ice storage cooling system in this embodiment may specifically include: S101. Obtain the real-time power data and weather condition data of the photovoltaic power station, collect the photovoltaic output power P(t) and sunshine intensity I(t) per minute through sensors, and use time series analysis method to calculate the power volatility F(t)=|P(t)-P(t - 1)| / P(t - 1) to obtain the quantization result of power volatility.

[0015] Collect the photovoltaic output power P(t) and sunshine intensity I(t) per minute through sensors, where P(t) is the power at the current time point and P(t - 1) is the power of the previous minute. Calculate the power volatility F(t) = |P(t) - P(t - 1)| / P(t - 1). If P(t - 1) is 0 and P(t) is non - zero, then F(t) is set to 0. Obtain the sunshine intensity I(t) in the time series and calculate the correlation coefficient between the power volatility F(t) and the sunshine intensity I(t) using the Pearson correlation coefficient. If the power volatility F(t) exceeds the preset threshold statistically analyzed based on historical data, then the fluctuation is judged as abnormal. Process the time series by the sliding window method with a window size of 5 minutes and a step size of 1 minute, calculate the average power volatility and average sunshine intensity for 5 consecutive minutes to obtain the smoothed fluctuation trend and average sunshine intensity. Calculate the correlation coefficient between the smoothed fluctuation trend and the average sunshine intensity using the Pearson correlation coefficient to determine the relationship between the volatility and weather conditions.

[0016] Exemplarily, by collecting data on photovoltaic output power and sunshine intensity per minute through sensors, the operating status of the photovoltaic system can be monitored and analyzed in real time. The power volatility, as an important indicator to measure the change of photovoltaic output power, can reflect the stability of the system operation, while the sunshine intensity is one of the main external factors affecting power fluctuations. The following will conduct a detailed analysis and give examples around technical topics such as power volatility calculation, Pearson correlation coefficient analysis, abnormal fluctuation judgment, and sliding window smoothing processing. As an implementation method, the calculation of power volatility F(t) can intuitively reflect the dynamic change of photovoltaic power. For example, assume that on a sunny morning, the power P(t - 1) in a certain minute is 100 kW, and the power P(t) in the next minute is 110 kW. Then F(t) = |110 - 100| / 100 = 0.1, indicating that the power volatility is 10%. If the power P(t - 1) in a certain minute is 0 kW and the power P(t) is 50 kW, then according to the definition, F(t) is set to 0. This calculation method can effectively capture the sudden change of power from zero to non - zero, such as the scenario when the photovoltaic system just starts at sunrise in the early morning. Through this method, basic data can be provided for subsequent fluctuation analysis. Further, calculating the correlation between the power volatility F(t) and the sunshine intensity I(t) using the Pearson correlation coefficient can reveal the degree of association between the two. For example, assume that data for 60 minutes is collected on a certain day. During a certain period, the sunshine intensity I(t) gradually rises from 500 W / m² to 800 W / m², and the power volatility F(t) also shows a certain upward trend. Through calculation, the Pearson correlation coefficient between the two may be close to 0.7, indicating that the change in sunshine intensity has a strong positive impact on power fluctuations. This analysis helps to understand the role of weather conditions in the operation stability of the photovoltaic system and provides a basis for optimizing control strategies. Specifically, for the judgment of abnormal fluctuations, the preset threshold can be obtained through statistical analysis of historical data. Assume that through a large amount of historical data analysis, the average value of the power volatility F(t) under normal conditions is 0.15, and the standard deviation is 0.05. Then the threshold can be set to 0.25, that is, the mean plus two standard deviations. If the calculated value of F(t) in a certain minute is 0.3, exceeding the threshold, it is judged as an abnormal fluctuation. For example, on a cloudy day, the rapid movement of clouds causes a sharp change in sunshine intensity, and the power drops suddenly from 200 kW to 50 kW, and F(t) reaches 0.75, which is obviously abnormal. Such a judgment mechanism can timely detect potential problems, such as equipment failures or weather mutations, and provide early warnings for maintenance personnel. It should be noted that the introduction of the sliding window method can further smooth the data, reduce noise interference, and thus more clearly observe the fluctuation trend. For example, assume that the window size is 5 minutes and the step size is 1 minute. The power volatilities F(t) within a certain 5 - minute period are 0.1, 0.12, 0.15, 0.09, and 0.11 respectively. Then the average volatility within the window is (0.1 + 0.12 + 0.15 + 0.09 + 0.11) / 5 = 0.114.Meanwhile, calculate the average sunshine intensity within the calculation window. For example, the corresponding values are 600, 610, 590, 605, and 595 watts per square meter respectively, and the average value is 600 watts per square meter. Through this smoothing process, a more stable fluctuation trend and sunshine intensity change curve can be obtained. Further, calculate the Pearson correlation coefficient for the smoothed data again, which can more accurately reflect the relationship between the fluctuation trend and sunshine intensity. For example, if the smoothed volatility trend and the average sunshine intensity show a strong negative correlation, the correlation coefficient may be -0.6, indicating that the more stable the sunshine intensity, the smaller the power fluctuation. This analysis result can be used to optimize the power prediction model of the photovoltaic system and improve the prediction accuracy. For example, in practical applications, a certain photovoltaic power station found through the above method that during the period with frequent summer afternoon thunderstorms, the correlation between the smoothed fluctuation trend and sunshine intensity is low, and the volatility frequently exceeds the threshold. After further analysis in combination with meteorological data, the operation and maintenance team optimized the response strategy of the inverter, enabling it to better adapt to short-term weather changes and reducing the number of abnormal shutdowns.

[0017] S102. Extract eigenvalues from the power volatility F(t). For the operation log of the ice-making equipment, calculate the energy input E(t) per unit time = ∫P(t)dt, and use a convolutional neural network to process the instantaneous mapping relationship between F(t) and E(t) to determine the influence coefficient K(t) of the fluctuation on the ice-making energy input.

[0018] Process F(t) using a sliding window with a window size of 3 minutes and a step size of 1 minute to obtain a smoothed fluctuation sequence. Extract eigenvalues from the smoothed fluctuation sequence to form a fluctuation feature sequence. Calculate the energy per unit time E(t) = ∫P(t)dt through integration. Use a convolutional neural network to analyze the instantaneous mapping relationship between the fluctuation feature sequence and E(t) to obtain a mapping feature matrix containing the weights of the convolutional kernel and the output of the pooling layer, where the convolutional kernel size is 3×1, the step size is 1, and the activation function is ReLU. Calculate the influence coefficient K(t) for the mapping feature matrix. If K(t) exceeds the preset threshold of 2, it is determined that the influence of the fluctuation on the energy input is significant, and the mapping feature matrix is updated. Process the relationship between the smoothed fluctuation sequence and the energy per unit time E(t) through a random forest algorithm to obtain the energy fluctuation distribution. Adjust the influence coefficient K(t) according to the energy fluctuation distribution. If the deviation between the adjusted K(t) and the instantaneous mapping relationship is less than the preset threshold of 1, determine the dynamic influence of the fluctuation on the ice-making energy input. Conduct a joint analysis of the adjusted K(t) and the time series data through a convolutional neural network to obtain the final fluctuation influence coefficient sequence.

[0019] Exemplarily, as an implementation, processing F(t) using a sliding window can smooth the data and reduce noise interference. Assuming the window size is 3 minutes and the step size is 1 minute, and the power volatility within a certain 3 minutes is 0.1, 0.12, and 0.15 respectively, then the average volatility within the window is 0.123. This smoothing process can obtain a more stable fluctuation trend, which is helpful for subsequent analysis. Further, extracting eigenvalue from the smoothed fluctuation sequence to form a fluctuation feature sequence can capture key fluctuation patterns. For example, the maximum value, minimum value, and mean value within each window can be extracted as features, and these features can reflect the intensity and distribution of fluctuations. By integrating to calculate the unit time energy E(t), the actual output effect of the photovoltaic system can be evaluated. Assuming that within a certain 5 minutes, the power curve is approximately a parabola, the total energy output during this period can be obtained through integration. It should be noted that using a convolutional neural network to analyze the instantaneous mapping relationship between the fluctuation feature sequence and E(t) can capture complex non-linear relationships. The network may include multiple convolutional layers and pooling layers for extracting local and global features of the time series. Specifically, the convolutional neural network includes 3 convolutional layers, and the number of filters is 16 / 32 / 64 respectively. The pooling layer uses max pooling with a pooling size of 2×1. The obtained mapping feature matrix reflects the influence mode of fluctuations on energy output. As an implementation, calculating the influence coefficient K(t) for the mapping feature matrix can quantify the influence degree of fluctuations on energy input. If K(t) exceeds the preset threshold of 2, it indicates that the influence of fluctuations on energy input is significant. At this time, the mapping feature matrix needs to be updated to adapt to the new relationship mode. This dynamic adjustment mechanism can improve the adaptability and accuracy of the model. Specifically, by processing the relationship between the smoothed fluctuation sequence and the unit time energy E(t) using the random forest algorithm, the energy fluctuation distribution can be obtained. The advantage of the random forest is that it can handle non-linear relationships and multi-dimensional features, and is suitable for analyzing complex energy fluctuation patterns. According to the obtained distribution, the influence coefficient K(t) can be further adjusted to make it more in line with the actual situation. Further, using a convolutional neural network to jointly analyze the adjusted K(t) and the time series data can obtain the final fluctuation influence coefficient sequence. This method comprehensively considers the characteristics of the time series and the fluctuation influence, and can more accurately describe the dynamic influence of fluctuations on the ice-making energy input.

[0020] S103. Obtain the temperature sensor data T(t) and water volume sensor data V(t) in the ice storage tank, update the real-time energy input model E'(t)=K(t)×E(t) through the influence coefficient K(t), and calculate the dynamic changes of water temperature and water volume using the difference equations T'(t)=T(t - 1)+α×E'(t) and V'(t)=V(t - 1)+β×E'(t), where α and β are preset heat transfer and water storage coefficients.

[0021] Obtain the temperature data T(t) and water volume data V(t) in the ice storage tank, with the data acquisition frequency being once per minute. Calculate and update the energy E'(t)=K(t)×E(t) through the pre-set influence coefficient K(t), where E(t) is the current energy value and K(t) is the influence coefficient at time t. Generate the temperature sequence T'(t) and water volume sequence V'(t) using the difference equations T'(t)=T(t - 1)+α×E'(t) and V'(t)=V(t - 1)+β×E'(t), where α and β are the pre-determined response coefficients of temperature and water volume to energy respectively. Process the temperature sequence T'(t) and water volume sequence V'(t) through a sliding window, with the window size set to 5 minutes and the step size set to 1 minute, to obtain the smoothed temperature sequence and smoothed water volume sequence. Use linear regression analysis to analyze the mapping relationship between the smoothed temperature sequence and the updated energy E'(t), and combine with the time change t to obtain the temperature-energy correlation coefficient matrix. For the smoothed water volume sequence and the updated energy E'(t), use linear regression to process the dynamic relationship between the two, and combine with the time change t to obtain the water volume-energy correlation coefficient matrix. Combine the temperature-energy correlation coefficient matrix and the water volume-energy correlation coefficient matrix through matrix multiplication, and combine with the influence coefficient K(t) to obtain the comprehensive dynamic influence matrix. If the element value in the comprehensive dynamic influence matrix exceeds the preset threshold of 3, update the temperature sequence T'(t) and water volume sequence V'(t) through the difference equations to obtain the adjusted dynamic change sequence. According to the adjusted dynamic change sequence, use the weighted average method to fuse the temperature sequence T'(t), water volume sequence V'(t) and updated energy E'(t), with the weights determined according to the pre-determined contribution ratios of temperature and water volume to energy, to obtain the warm water energy distribution sequence in the ice storage tank.

[0022] Exemplarily, for the acquisition and processing of temperature and water volume data in the ice storage tank, by adjusting the energy input through a preset influence coefficient, the actual operating state of the system can be more accurately reflected. As an implementation method, the temperature T(t) and water volume V(t) data can be obtained at a collection frequency of once per minute. This high-frequency collection helps to capture the instantaneous changes of the system. Specifically, the energy E(t) is corrected by the influence coefficient K(t) to obtain the updated energy E'(t). This correction mechanism can dynamically adjust the energy input to adapt to the system requirements at different times. For example, during the peak ice-making period, K(t) may be relatively large, resulting in an increase in E'(t), thereby improving the ice-making efficiency. Further, difference equations are used to generate the temperature sequence T'(t) and the water volume sequence V'(t). This method can simulate the dynamic response of the system. The response coefficients α and β of temperature and water volume to energy reflect the physical characteristics of the system. As an implementation method, these coefficients can be determined through experiments. For example, α may be 0.02 °C / kWh and β may be 0.05 L / kWh. It should be noted that using a sliding window process can smooth data fluctuations and improve the reliability of the data. The window size is set to 5 minutes and the step size is 1 minute. This setting can not only retain sufficient detailed information but also effectively reduce noise interference. Linear regression analysis is used to explore the relationship between temperature, water volume, and energy. This method can quantify the mutual influence between parameters. For example, it may be found that for every 1 °C increase in temperature, the energy consumption increases by 2 kWh. By fusing the influences of temperature and water volume through matrix operations, a comprehensive dynamic influence matrix can be obtained. As an implementation method, the threshold can be set to 3. When the matrix element value exceeds this threshold, it indicates that a significant change has occurred in the system state and timely adjustment is required. This mechanism can improve the response speed and adaptability of the system. Finally, the temperature, water volume, and energy data are fused through the weighted average method to obtain the warm water energy distribution sequence in the ice storage tank. This method that comprehensively considers multiple factors can more comprehensively describe the system state. For example, the weights of temperature, water volume, and energy can be set to 0.3, 0.3, and 0.4 respectively to balance the influences of various factors. Through this data processing and analysis method, precise control and optimized operation of the ice-making system can be achieved, improving energy utilization efficiency and reducing operating costs.

[0023] S104. According to the dynamic change data of the water temperature T'(t) and the water volume V'(t), construct a melting ice cooling efficiency function η(t) = γ × T'(t) + δ × V'(t), where γ and δ are preset efficiency weight coefficients, and train a support vector machine model through historical cooling data to obtain the predicted value of η(t).

[0024] Obtain the dynamic water temperature sequence T'(t) and the dynamic water volume sequence V'(t) in the ice storage tank. Through the preset temperature efficiency weight coefficient γ and the water volume efficiency weight coefficient δ, calculate the efficiency function η(t)=γ×T'(t)+δ×V'(t) to obtain the initial cooling efficiency sequence η(t). Input the initial cooling efficiency sequence η(t) and historical data into the support vector machine model for training to obtain the efficiency prediction sequence η’(t). Perform a sliding window process on the efficiency prediction sequence η’(t), with a window size of 10 minutes and a step size of 1 minute, and use the mean smoothing method to obtain the smoothed efficiency sequence η’’(t). Conduct a linear regression analysis on the smoothed efficiency sequence η’’(t) and the dynamic water temperature sequence T'(t) by the least squares method to obtain the temperature efficiency correlation coefficient matrix A(t). Conduct a linear regression analysis on the smoothed efficiency sequence η’’(t) and the dynamic water volume sequence V'(t) by the least squares method to obtain the water volume efficiency correlation coefficient matrix B(t). Perform matrix weighted fusion on the temperature efficiency correlation coefficient matrix A(t) and the water volume efficiency correlation coefficient matrix B(t), with the weights determined by the preset temperature weight γ and water volume weight δ, to obtain the comprehensive efficiency influence matrix C(t). If the element value in the comprehensive efficiency influence matrix C(t) exceeds the preset threshold of 5, update the smoothed efficiency sequence through the difference equation η’’’(t)=η’’(t - 1)+λ×C(t) to obtain the adjusted efficiency sequence η’’’(t), where λ is the preset adjustment coefficient.

[0025] Exemplarily, by obtaining the dynamic sequence of water temperature T'(t) and the dynamic sequence of water volume V'(t) in the ice storage tank, the temperature efficiency weight coefficient γ and the water volume efficiency weight coefficient δ can be further introduced to construct the efficiency function η(t)=γ×T'(t)+δ×V'(t). As an implementation, assume γ is set to 0.4 and δ is set to 0.6, indicating that the influence of water volume on the cooling efficiency is slightly greater than that of temperature. Specifically, at a certain time point t, the water temperature T'(t) is 5°C and the water volume V'(t) is 1000 L, then η(t)=0.4×5 + 0.6×1000 = 602 units. This efficiency function can initially reflect the cooling capacity of the ice storage tank and provide basic data for subsequent optimization. It should be noted that the values of γ and δ can be determined through experiments or historical data analysis to ensure that they accurately reflect the physical characteristics of the system. Specifically, the efficiency weight coefficients are determined by multiple linear regression of historical cooling data, where the temperature weight γ ranges from 0.32 to 0.45 and the water volume weight ranges from 0.55 to 0.68; this helps to improve the scientific nature of efficiency evaluation and lay a foundation for the operation optimization of the ice-making system. Further, the initial cooling efficiency sequence η(t) and historical data are input into the support vector machine model for training to obtain the efficiency prediction sequence η’(t). For example, assume that the historical data includes temperature and water volume records every minute in the past week. The support vector machine can learn the patterns of these data and predict the efficiency trend in the next hour. Specifically, during the peak ice-making period, the model may predict that η’(t) rises from 600 units to 650 units, reflecting the efficiency improvement brought by the increased energy input. The advantage of this prediction method is that it can detect the efficiency change trend in advance, gain time for system adjustment, and thus avoid energy waste or insufficient cooling. The sliding window processing is adopted for the efficiency prediction sequence η’(t), the window size is 10 minutes, the step size is 1 minute, and the mean smoothing method is used to obtain the smoothed efficiency sequence η’’(t). As an implementation, assume that a certain segment of η’(t) data is 600, 610, 620, 615, 605, 595, 590, 600, 610, 620. The mean of the first 10 values of the sliding window is (600 + 610 + 620 + 615 + 605 + 595 + 590 + 600 + 610 + 620) / 10 = 606.5 units. This can effectively reduce the efficiency fluctuation caused by sensor noise and further improve the reliability of the data. Specifically, this smoothing process can help the operator more clearly identify the long-term trend of efficiency and avoid misjudging the system state due to short-term fluctuations. Through the least squares method, a linear regression analysis is performed on the smoothed efficiency sequence η’’(t) and the dynamic sequence of water temperature T'(t) to obtain the temperature efficiency correlation coefficient matrix A(t). For example, it is found through analysis that for every 1°C increase in water temperature, the efficiency may increase by 10 units, indicating that temperature has a positive impact on efficiency.As an implementation method, this relationship can be verified through multiple experiments to ensure that matrix A(t) accurately reflects the contribution of temperature changes. Similarly, a regression analysis is performed on η’’(t) and the water volume dynamic sequence V'(t) to obtain the water volume efficiency correlation coefficient matrix B(t). Assuming that the efficiency increases by 15 units for every 100L increase in water volume, B(t) can quantify this impact. The advantage of this analysis method is that it can clarify the respective effects of temperature and water volume on efficiency, providing data support for subsequent optimization. The temperature efficiency correlation coefficient matrix A(t) and the water volume efficiency correlation coefficient matrix B(t) are matrix-weighted and fused to generate the comprehensive efficiency impact matrix C(t). Specifically, with preset weights γ = 0.4 and δ = 0.6 for weighting, C(t) = 0.4×A(t) + 0.6×B(t). For example, if a certain element value of A(t) is 10 and the corresponding element value of B(t) is 15, then the value of this element in C(t) is 0.4×10 + 0.6×15 = 13. This fusion method comprehensively considers the synergistic effect of temperature and water volume, making the system state assessment more comprehensive. It should be noted that if the element value in C(t) exceeds the threshold of 5, for example, reaching 13, it indicates that the efficiency is significantly affected and the operation strategy needs to be adjusted. If the element value of C(t) exceeds the threshold of 5, the smoothed efficiency sequence is updated through the difference equation η’’’(t) = η’’(t - 1) + λ×C(t) to obtain the adjusted efficiency sequence η’’’(t). As an implementation method, the adjustment coefficient λ is set to 0.1. Assuming η’’(t - 1) is 600 units and C(t) is 13, then η’’’(t) = 600 + 0.1×13 = 601.3 units. This adjustment mechanism can quickly respond to system changes, such as increasing the efficiency target value when the ice-making demand suddenly increases. Specifically, this method can enhance the adaptability of the system and ensure that the cooling efficiency is always in the best state. Further, by continuously monitoring η’’’(t), it can provide a basis for equipment maintenance. For example, finding that the efficiency decreases continuously over a long period may be a signal of the aging of the ice storage tank. For example, in actual operation, a certain ice storage tank increases the water volume by increasing the energy input during the low electricity price period at night. T'(t) rises from 4°C to 6°C, V'(t) increases from 800L to 1200L, η(t) increases from 500 to 730, and η’(t) predicts that the efficiency during the peak period the next day can reach 750 units. After smoothing and adjustment, η’’’(t) stabilizes at 740 units. This multi-step process from data collection to efficiency optimization significantly improves the energy utilization efficiency, reduces the operation cost, and provides reliable support for system prediction and maintenance.

[0026] S105. Obtain the building load demand data L(t) and the ambient temperature data A(t). Analyze the time series characteristics of L(t) using a long short-term memory network. Combine A(t) to calculate the load demand change rate R(t) = |L(t) - L(t - 1)| / L(t - 1), and judge the dynamic coupling characteristics of the load demand.

[0027] Obtain the dynamic sequence Lt of building load demand and the dynamic sequence At of ambient temperature. Perform time series decomposition on Lt using a long short-term memory network implemented by TensorFlow to obtain the trend component T_Lt and the periodic component C_Lt of the load demand. Calculate the load demand change rate sequence Rt through the formula Rt = |Lt - Lt - 1| / max(Lt - 1, 1), where the denominator takes the maximum value of Lt - 1 and 1 to avoid the denominator being zero. Combine the dynamic sequence At of ambient temperature, and use the Pearson correlation coefficient in the Scipy library to analyze the correlation between Rt and At to obtain the correlation coefficient sequence Pt. For the correlation coefficient sequence Pt, use the rolling function of the Pandas library for sliding window processing, with a window size of 15 minutes and a step size of 2 minutes, and obtain the smoothed correlation coefficient sequence P’t through the mean smoothing method. According to the smoothed correlation coefficient sequence P’t, use a random forest model implemented by Scikit-learn to fit the non-linear relationship between the load demand change rate Rt and the ambient temperature At to obtain the fitting relationship function Ft. If the weight values of the variables in the fitting relationship function Ft are greater than the preset threshold 8, update the change rate sequence through the difference equation R’t = Rt - 1 + μ × Ft to obtain the adjusted change rate sequence R’t, where μ is the preset adjustment coefficient. Perform prediction on the adjusted change rate sequence R’t using a long short-term memory network implemented by TensorFlow, and combine the dynamic sequence At of ambient temperature to obtain the load demand prediction sequence L’t for the next time period. According to the load demand prediction sequence L’t and the trend component T_Lt, use a weighted fusion method, with the weight determined by P’t, to obtain the comprehensive load demand dynamic sequence L’’t.

[0028] Exemplarily, by obtaining the dynamic sequence of building load demand Lt and the dynamic sequence of ambient temperature At, key input data can be provided for the operation optimization of the ice-making system or the cooling system. As an implementation, the dynamic sequence of building load demand Lt can be understood as the electricity demand for cooling per hour in a certain office building. For example, it may reach 500 kW during the peak period at noon in summer, and drop to 100 kW during the low valley period at night. The dynamic sequence of ambient temperature At records the changes in outdoor temperature. For example, it rises from 25°C in the morning to 35°C in the afternoon. Using the long short-term memory network implemented by TensorFlow to perform time series decomposition on Lt can extract the trend component T_Lt and the periodic component C_Lt. Specifically, the trend component T_Lt reflects the long-term changes in load demand. For example, the trend of gradually increasing load on weekdays, while the periodic component C_Lt captures the regular fluctuations of the morning and evening peaks every day. The advantage of this decomposition is that it splits the complex load data into interpretable parts, laying a foundation for subsequent prediction and optimization. Further, by calculating the load demand change rate sequence Rt through the formula Rt = |Lt - Lt-1| / max(Lt-1, 1), the dynamic changes in the load can be quantified. For example, if Lt-1 was 400 kW in the previous hour and the current Lt is 450 kW, then Rt = |450 - 400| / 400 = 0.125, indicating that the load has increased by 12.5%. Taking the maximum value of Lt-1 and 1 in the denominator avoids calculation errors when the load is zero. As an implementation, when the load is extremely low at night, for example, Lt-1 is 0 kW and Lt is 50 kW, then Rt = 50 / 1 = 50, reflecting the drastic change of suddenly starting equipment. This change rate sequence can help operators quickly identify the intensity of load fluctuations. Combining the dynamic sequence of ambient temperature At and using the Pearson correlation coefficient in the Scipy library to analyze the correlation between Rt and At, the correlation coefficient sequence Pt can be obtained. For example, in the records of a certain day, when At rises from 30°C to 35°C, Rt increases from 0.1 to 0.15, and the calculated Pt may be 0.85, indicating a strong positive correlation between the temperature increase and the load change rate. Specifically, this analysis can reveal how the ambient temperature drives the electricity demand and provide a basis for the energy distribution of the cooling system. It should be noted that the positive or negative value of Pt can also indicate whether the temperature change increases or decreases the load fluctuation. For the correlation coefficient sequence Pt, the rolling function of the Pandas library is used for sliding window processing. The window size is 15 minutes and the step size is 2 minutes, and P’t is obtained through mean smoothing. For example, a certain section of Pt data is 0.8, 0.85, 0.9, 0.87, 0.82, and the first value of P’t after smoothing is (0.8 + 0.85 + 0.9 + 0.87 + 0.82) / 5 = 0.848. This smoothing process can reduce the noise caused by short-term temperature fluctuations and make the correlation trend clearer. As an implementation, operators can judge whether the long-term impact of temperature on the load is stable based on this.According to P’t, the random forest model implemented using Scikit-learn is used to fit the non-linear relationship between Rt and At, and the fitting relationship function Ft is obtained. For example, the model may find that when At exceeds 32°C, the growth rate of Rt significantly accelerates, and the weight value may reach 10, exceeding the threshold of 8. Further, if the variable weight in Ft is greater than 8, the change rate sequence is updated through the difference equation R’t = Rt-1 + μ × Ft. Assuming μ is 0.05, Rt-1 is 0.12, and Ft is 10, then R’t = 0.12 + 0.05 × 10 = 0.17. This adjustment can quickly respond to temperature-driven load changes and improve the prediction accuracy. The long short-term memory network of TensorFlow is used to predict R’t. Combining At, the load demand prediction sequence L’t for the next time period can be obtained. For example, when predicting that At will be 36°C at noon the next day, L’t may increase from the current 480 kW to 520 kW. Specifically, this prediction method can plan the operation time of the ice-making system in advance and avoid insufficient cooling supply. According to L’t and T_Lt, P’t is used as the weight for weighted fusion to generate the comprehensive load demand dynamic sequence L’’t. For example, if P’t is 0.85, T_Lt is 490 kW, and L’t is 520 kW, then L’’t = 0.85 × 520 + (1 - 0.85) × 490 = 515.5 kW. This fusion method comprehensively considers the trend and prediction results, making the load demand assessment more comprehensive. It should be noted that this multi-step analysis from data collection to prediction optimization can significantly improve the energy efficiency of the cooling system. Further, continuously monitoring L’’t can also provide clues for equipment maintenance. For example, finding that the long-term load prediction is too high may be a signal of a decrease in ice-making efficiency. The advantage of this method is that it enhances the adaptability and reliability of the system through a data-driven approach.

[0029] S106. Through the ice melting cooling efficiency η(t) and the load demand change rate R(t), a dynamic coupling relationship model C(t) = ω × η(t) + ψ × R(t) is constructed, where ω and ψ are preset coupling weights. If C(t) exceeds the preset threshold, the time-delay effect characteristic τ(t) = argmax(C(t) - C(t - τ)) is calculated by the sliding window method to determine the energy transfer delay between ice making and cooling.

[0030] By collecting the time series data of the ice melting cooling efficiency η(t) and the load demand change rate R(t), the dynamic coupling relationship sequence is calculated using the formula C(t)=ω×η(t)+ψ×R(t), where ω and ψ are preset coupling weights, and the coupling model sequence C(t) is obtained. If C(t) is greater than the preset threshold, the C(t) sequence is processed by the sliding window method with a window size of 20 minutes and a step size of 5 minutes. When calculating the time-delay effect feature τ(t)=argmax_{τ∈[5,30]}(C(t)-C(t-τ)), the time delay τ is determined, and the time-delay feature sequence τ(t) is obtained, where the unit of τ is minutes. Using the support vector machine algorithm in Scikit-learn and selecting the radial basis function as the kernel function, C(t) and τ(t) are used as inputs for non-linear mapping processing to obtain the mapping relationship function M(t). Through the long short-term memory network implemented by TensorFlow, the mapping relationship function M(t) and the ice melting cooling efficiency η(t) are used as inputs for joint time series analysis to predict the coupling model sequence C’(t) in the next time period. According to the predicted C’(t) and the time-delay feature sequence τ(t), the adjusted coupling sequence C’’(t) is calculated using the weighted fusion method, and the weight is determined by the product of M(t) and C’(t), obtaining the adjusted coupling sequence C’’(t). For the adjusted coupling sequence C’’(t), the load demand change rate is updated through the difference equation R’(t)=R(t-1)+λ×C’’(t), where λ is the preset adjustment coefficient, obtaining the adjusted change rate sequence R’(t). Using the random forest model implemented by Scikit-learn, the adjusted change rate sequence R’(t) and the time delay τ are used as inputs for regression analysis to obtain the fitting relationship function F(t) between the load demand change rate and the energy transfer delay, determining the final dynamic coupling characteristics.

[0031] Exemplarily, by collecting the time series data of the ice melting and cooling efficiency η(t) and the load demand change rate R(t), a basis for dynamic optimization of the cooling system can be provided. For example, in the ice making and cooling system of an office building, η(t) represents the energy conversion efficiency of the ice melting process, which may reach 0.9 when making ice during the low load period at night, and drop to 0.75 during the peak period during the day due to the increase in the operating load of the equipment. R(t) reflects the load demand change rate. For example, due to the sharp increase in air conditioner use at noon, it rises from 0.1 to 0.15. The dynamic coupling relationship sequence is calculated using the formula C(t)=ω×η(t)+ψ×R(t). Assuming ω is 0.6 and ψ is 0.4, when η(t) is 0.8 and R(t) is 0.12 at a certain moment, C(t)=0.6×0.8 + 0.4×0.12 = 0.528. As an implementation, C(t) quantifies the combined effect of efficiency and load changes. If it is greater than the preset threshold of 0.5, it indicates that the system needs to pay attention to dynamic adjustment. Specifically, when C(t) exceeds the threshold, the sliding window method is used to process the C(t) sequence. The window size is 20 minutes and the step size is 5 minutes, and the time lag effect feature τ(t) is calculated. For example, on a certain day, C(t) slowly rises from 0.52 to 0.58. By comparing the data before and after, it is found that the maximum difference appears 15 minutes ago, so τ(t) is 15 minutes. This reflects the delay in the impact of load changes on the cooling efficiency. Further, using the support vector machine algorithm, with C(t) and τ(t) as inputs and selecting the radial basis function kernel, non-linear relationships can be captured. For example, when τ(t) is 10 minutes and C(t) is 0.55, the mapping function M(t) may output 0.62, indicating the potential association strength between the two. The advantage of this method is to reveal how time delay affects the system response. As an implementation, through the long short-term memory network to analyze M(t) and η(t), the C’(t) in the next time period is predicted. For example, when the current M(t) is 0.6 and η(t) is 0.78, the network may predict C’(t) to be 0.59. This can judge the system coupling state in advance and help the operator adjust the ice making strategy. Combining C’(t) and τ(t), the weighted fusion is used to calculate C’’(t), and the weight is determined by the product of M(t) and C’(t). For example, when M(t) is 0.6 and C’(t) is 0.59, the weight is 0.354, and C’’(t) may be 0.57 after fusion. This adjustment makes the prediction closer to the actual operating state and improves the reliability. Further, the change rate is updated through the difference equation R’(t)=R(t - 1)+λ×C’’(t). Assuming λ is 0.05, R(t - 1) is 0.13 and C’’(t) is 0.57, then R’(t)=0.13 + 0.05×0.57 = 0.1585. This reflects the dynamic impact of the coupling effect on the load change. It should be noted that this update can quickly respond to system changes and avoid insufficient cooling.For example, at noon on a certain day, the temperature suddenly rises, and R’(t) increases from 0.12 to 0.16, indicating that the ice melting amount needs to be increased. Specifically, the random forest model is used to analyze R’(t) and τ(t) to obtain the fitting relationship function F(t). For example, when τ(t) is 12 minutes and R’(t) is 0.15, F(t) may output 0.18, indicating that the delay is positively correlated with the change rate. This can help the operator understand the impact of energy transfer delay on the load. For example, during the summer peak period, when τ(t) is long, F(t) indicates that the cooling equipment needs to be started in advance. As an implementation method, if F(t) shows that τ(t) exceeds 15 minutes and R’(t) grows rapidly, the ice making amount can be increased by 20 minutes in advance to ensure stable cooling. This multi-step analysis progresses layer by layer from efficiency, change rate to time delay characteristics to ensure accurate energy distribution.

[0032] S107. Extract full-cycle features from the dynamic coupling relationship C(t) and the time-delay effect τ(t), and use the random forest algorithm to perform multivariate interactive regression analysis on T'(t), V'(t), η(t), and R(t) to obtain the energy efficiency prediction value Q(t) of the system at different time scales.

[0033] By dynamically coupling C(t), time-delay effect τ(t), ice melting efficiency η(t), and load change R(t) through full-cycle data acquisition, preprocessing methods are used to extract the periodic feature T'(t) and velocity feature V'(t) to obtain the initial feature sequence. According to the initial feature sequence, the random forest algorithm is used to perform multivariate regression processing on the periodic feature T'(t), velocity feature V'(t), ice melting efficiency η(t), and load change R(t) to obtain the preliminary energy efficiency prediction sequence Q1(t). For the preliminary energy efficiency prediction sequence Q1(t), the fluctuation feature sequence F(t) of Q1(t) within the time scale is calculated by the sliding window method, with the window size set to 30 minutes and the step size set to 10 minutes, to obtain the fluctuation feature sequence F(t). If the fluctuation feature sequence F(t) exceeds the preset threshold, the support vector machine algorithm is used to perform non-linear mapping processing on F(t) and the time-delay effect τ(t), and the radial basis function is selected as the kernel function to obtain the mapping feature sequence M(t). Through the long short-term memory network, joint time series analysis is performed on the mapping feature sequence M(t), dynamic coupling C(t), and ice melting efficiency η(t) to predict the energy efficiency prediction sequence Q2(t) for the next time period. According to the predicted energy efficiency prediction sequence Q2(t) and the fluctuation feature sequence F(t), a weighted fusion method is used to calculate the adjusted energy efficiency sequence Q3(t), and the weight is determined by the product of M(t) and Q2(t), to obtain the adjusted energy efficiency sequence Q3(t). The load change sequence R'(t) = R(t - 1) + λ × Q3(t) is updated through the difference equation, where λ is the preset adjustment coefficient, and the random forest algorithm is used to perform regression analysis on R'(t), time-delay effect τ(t), and time scale to obtain the final energy efficiency prediction sequence Q(t).

[0034] Exemplarily, by dynamically coupling C(t), time-delay effect τ(t), ice melting efficiency η(t), and load change R(t) through full-cycle data acquisition, a dynamic optimization data basis can be provided for the building energy system. As an implementation, the preprocessing method extracts the periodic feature T'(t) and the velocity feature V'(t) with the aim of capturing the periodic pattern and change rate of the system operation. For example, in the ice-making and cooling system of an office building, data is collected for 24 hours a day, and it is found that η(t) is relatively high at night, about 0.9, because the ice-making efficiency is optimal at the low load trough, while during the day, due to the increase in air-conditioning load, η(t) drops to 0.7. T'(t) may show that 24 hours is a significant cycle, while V'(t) reflects the decline rate of η(t) from night to day, such as a decline of 0.05 per hour. This feature extraction lays the foundation for subsequent analysis and helps to reveal the dynamic relationship between efficiency and load. Specifically, based on the initial feature sequence, the random forest algorithm performs multivariate regression on T'(t), V'(t), η(t), and R(t) to generate the preliminary energy efficiency prediction sequence Q1(t). The advantage of the random forest lies in dealing with the complex relationships between multiple variables. For example, during the peak summer period, T'(t) shows that the load demand fluctuates every 12 hours, V'(t) indicates an accelerated change rate, R(t) rises from 0.1 to 0.15, and η(t) is 0.75. The random forest combines these factors and may predict Q1(t) to be 0.65, representing the current energy efficiency state. The benefit of this step is to provide a preliminary prediction and lay the foundation for subsequent refined analysis. Further, for Q1(t), the sliding window method calculates the fluctuation feature sequence F(t) with a window size of 30 minutes and a step size of 10 minutes to quantify the short-term fluctuations of energy efficiency. For example, at noon on a certain day, Q1(t) rises from 0.62 to 0.68, and after calculating F(t), it shows that the fluctuation amplitude is 0.06. If F(t) exceeds the preset threshold of 0.05, it indicates that the energy efficiency of the system fluctuates significantly and further processing is required. This reflects the impact of load mutation on energy efficiency and helps the operator to detect abnormalities in a timely manner. As an implementation, if F(t) exceeds the threshold, the support vector machine uses the radial basis function kernel to perform a non-linear mapping on F(t) and τ(t) to generate the mapping feature sequence M(t). For example, τ(t) is 15 minutes and F(t) is 0.06, and M(t) may output 0.7, revealing the potential association between fluctuations and time delay. This method can capture non-linear features, improve the prediction accuracy, and provide support for subsequent time series analysis. By jointly analyzing M(t), C(t), and η(t) through the long short-term memory network, the energy efficiency sequence Q2(t) for the next time period is predicted. Specifically, assuming that the current M(t) is 0.7, C(t) is 0.58, and η(t) is 0.76, the network may predict Q2(t) to be 0.61. This step utilizes the time dependence of historical data to predict the future energy efficiency trend, which helps to adjust the ice-making strategy in advance to avoid insufficient cooling.According to Q2(t) and F(t), the adjusted energy efficiency sequence Q3(t) is calculated by weighted fusion, and the weight is determined by the product of M(t) and Q2(t). For example, when M(t) is 0.7 and Q2(t) is 0.61, the weight is 0.427, and after fusion, Q3(t) may be 0.6. This adjustment makes the prediction closer to the actual operating state, improves reliability, especially in scenarios with large load fluctuations. Further, the load change sequence is updated through the difference equation R'(t) = R(t - 1) + λ × Q3(t), where λ is set to 0.05. For example, when R(t - 1) is 0.13 and Q3(t) is 0.6, R'(t) is updated to 0.16. This reflects the dynamic impact of energy efficiency on load changes and ensures that the system can quickly respond to demand changes. It should be noted that random forest performs regression analysis on R'(t), τ(t), and the time scale to generate the final energy efficiency prediction sequence Q(t). For example, when τ(t) is 12 minutes and R'(t) is 0.15, Q(t) may be 0.67. This multi-variable regression can comprehensively consider the effects of time delay and load changes. For example, on a high-temperature day in summer, τ(t) is relatively long and R'(t) grows rapidly, and Q(t) indicates that the equipment needs to be started in advance to ensure stable cooling. This progressive analysis, from feature extraction to prediction optimization, not only reduces energy consumption but also improves the system's adaptability, providing a scientific basis for energy management.

[0035] S108. Obtain the prediction result of Q(t). For the historical data in extreme weather scenarios, if the deviation |D(t)| = |Q(t) - Q'(t)| between Q(t) and the actual energy efficiency Q'(t) exceeds the preset threshold, the model parameters are adjusted through adaptive Kalman filtering to update the full-cycle energy efficiency evaluation framework.

[0036] Obtain the predicted energy efficiency sequence Q(t) and the actual energy efficiency sequence Q'(t) through historical data in extreme weather scenarios, calculate the deviation sequence |D(t)| = |Q(t) - Q'(t)|, and obtain the distribution characteristics of the deviation sequence |D(t)|. If the deviation sequence |D(t)| exceeds the preset threshold, the adaptive Kalman filter is used to adjust the model parameters. Combining the full-cycle feature sequence T'(t) and the dynamic coupling relationship C(t) in the historical data, the updated model parameter sequence P(t) is obtained. According to the updated model parameter sequence P(t), recalculate the nonlinear mapping characteristics of the time-delay effect sequence τ(t) and the load change sequence R(t). Use the support vector machine algorithm to process the full-cycle feature sequence T'(t), select the radial basis function as the kernel function, and obtain the mapping feature sequence M'(t). Through the mapping feature sequence M'(t) and the dynamic coupling relationship C(t), use the long short-term memory network to perform time series prediction on the time-delay effect sequence τ(t), and obtain the time-delay effect prediction sequence τ’(t) for the next time period. Obtain the time-delay effect prediction sequence τ’(t) and the load change sequence R(t). Through the random forest algorithm, perform multivariate regression analysis on the full-cycle feature sequence T'(t), the mapping feature sequence M'(t), and the time-delay effect prediction sequence τ’(t), and obtain the adjusted load change prediction sequence R'(t). For the adjusted load change prediction sequence R'(t), calculate the fluctuation feature sequence F'(t) within the time scale through the sliding window method. The window size is set to 30 minutes and the step size is set to 10 minutes, and the fluctuation feature sequence F'(t) is obtained. According to the fluctuation feature sequence F'(t) and the adjusted load change prediction sequence R'(t), use the weighted fusion method to calculate the final energy efficiency prediction sequence Q'(t). The weight is determined by the product of the mapping feature sequence M'(t) and the time-delay effect prediction sequence τ’(t), and the final energy efficiency prediction sequence Q'(t) is obtained.

[0037] Exemplarily, the predicted energy efficiency sequence Q(t) and the actual energy efficiency sequence Q'(t) are obtained through historical data in extreme weather scenarios, and the deviation sequence |D(t)| = |Q(t) - Q'(t)| is calculated to evaluate the prediction accuracy of the model under abnormal conditions. For example, in the typhoon weather in a certain city in summer, the historical data records the operating status of the building cooling system. The predicted energy efficiency Q(t) is 0.68, while the actual energy efficiency Q'(t) is 0.62, and the deviation |D(t)| is 0.06. Further, by statistically analyzing multi-day data, the deviation sequence |D(t)| exhibits a normal distribution characteristic, with a mean of approximately 0.05 and a standard deviation of 0.02. If the preset threshold is 0.04, then the deviation exceeds the threshold in some periods, indicating that the model prediction under extreme weather has deviations and needs to be optimized. As an implementation manner, when the deviation sequence |D(t)| exceeds the threshold, the adaptive Kalman filter dynamically adjusts the model parameters, combines the full-cycle feature sequence T'(t) and the dynamic coupling relationship C(t), and generates the updated parameter sequence P(t). Specifically, in the above typhoon scenario, T'(t) shows that the load fluctuation intensifies within a 24-hour cycle, and C(t) reflects that the coupling strength between devices decreases from 0.55 to 0.45. The filtering algorithm adjusts the parameters according to these features to make P(t) more adaptable to the dynamic changes under extreme conditions. The advantage of this method is to improve the robustness of the model to the mutation environment, for example, to avoid inaccurate predictions caused by sudden increases in load. It should be noted that the updated parameter sequence P(t) is used to recalculate the nonlinear mapping characteristics of the time-delay effect sequence τ(t) and the load change sequence R(t). The support vector machine processes T'(t) with a radial basis function kernel to generate the mapping feature sequence M'(t). For example, in a certain hospital on a rainy and stormy day, T'(t) shows a 12-hour cycle, τ(t) extends from 10 minutes to 15 minutes, and the support vector machine outputs M'(t) as 0.72, revealing the nonlinear association between the time delay and the cycle. This mapping can capture complex relationships and improve the accuracy of subsequent predictions. Further, through M'(t) and C(t), the long short-term memory network performs time series prediction on τ(t) to obtain the time-delay effect prediction sequence τ’(t) for the next time period. For example, in the heavy snow weather, M'(t) is 0.68, C(t) is 0.50, and the network predicts that τ’(t) increases from 12 minutes to 18 minutes. This prediction utilizes the historical time dependence, which helps to adjust the cooling strategy in advance to avoid energy efficiency degradation caused by the extension of the time delay. Specifically, after obtaining τ’(t) and R(t), the random forest algorithm performs multivariate regression on T'(t), M'(t), τ’(t) to generate the adjusted load change prediction sequence R'(t). For example, in the hot and humid weather, T'(t) shows a 6-hour fluctuation, M'(t) is 0.70, τ’(t) is 14 minutes, and R(t) increases from 0.12 to 0.16. The random forest predicts R'(t) as 0.17. This method synthesizes the influence of multiple variables to ensure that the load prediction is close to the actual operating state.As an implementation, for R'(t), the sliding window method calculates the fluctuation feature sequence F'(t) with a 30 - minute window and a 10 - minute step. For example, in a commercial building during a thunderstorm, R'(t) rises from 0.15 to 0.18, and the calculated fluctuation amplitude of F'(t) is 0.03. If the threshold is 0.05, the fluctuation is controllable; if it exceeds the threshold, the operator is prompted to pay attention to the load mutation. This method of quantifying short - term fluctuations helps to detect anomalies in a timely manner and optimize the system response. Further, based on F'(t) and R'(t), the weighted fusion method calculates the final energy efficiency prediction sequence Q'(t), and the weight is determined by the product of M'(t) and τ’(t). For example, M'(t) is 0.72, τ’(t) is 15 minutes, the weight is 0.72×15 / 60 = 0.18, and after fusion, Q'(t) is 0.65. This adjustment makes the prediction closer to the actual operating state, especially effective in extreme weather. For example, in an office building during a snowstorm, Q'(t) indicates that the energy efficiency drops to 0.60, and the operator increases the ice - making capacity in advance based on this to ensure stable cooling supply. This progressive analysis, from deviation assessment to parameter optimization, and then to multi - variable prediction, not only improves the prediction accuracy but also enhances the system's adaptability to extreme conditions, providing reliable support for energy management.

[0038] It also includes an energy efficiency evaluation and optimization system for a photovoltaic - powered chilled - water storage system for implementing any one of the technical solutions described in steps S101 - S108, characterized by including: A data acquisition module for obtaining real - time power data, weather condition data, ice storage tank temperature data, water volume data, building load demand data, and ambient temperature data of a photovoltaic power station through a sensor network; A power fluctuation analysis module for calculating a power volatility sequence based on a sliding window algorithm and a Pearson correlation coefficient, and generating a quantification result of power volatility; A dynamic modeling module for processing a smoothed fluctuation feature sequence through a convolutional neural network and a random forest algorithm, generating a real - time energy input model, and outputting dynamic change parameters of a temperature sequence and a water volume sequence; A coupling relationship construction module for integrating a support vector machine model and a long short - term memory network, processing the time - delay effect of a chilled - water supply efficiency sequence and a load demand change rate sequence, and generating a dynamic coupling relationship sequence and an energy transfer delay sequence; An energy efficiency prediction engine module for performing multi - variable regression analysis and fluctuation feature mapping processing, and generating a full - cycle energy efficiency prediction sequence based on a comprehensive dynamic influence matrix and periodic feature extraction; A feedback optimization module for adaptively calibrating an influence coefficient sequence and a dynamic coupling relationship sequence according to a preset threshold, and updating system evaluation parameters through a matrix weighted fusion algorithm.

[0039] The specific embodiments described above further elaborate on the objective, technical solution and beneficial effects of the present invention. It should be understood that the above description is only the specific embodiments of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for optimizing energy efficiency evaluation of a photovoltaic power generation cold storage system, characterized in that: include: Obtain real-time power data and weather condition data of photovoltaic power stations and calculate power fluctuation rate series; Extracting characteristic values ​​from the power fluctuation rate sequence to determine a sequence of influence coefficients of fluctuations on ice-making energy input; Acquire temperature data and water volume data in the ice storage tank, update the real-time energy input model through the influence coefficient sequence, and calculate the dynamic changes of the temperature sequence and the water volume sequence; Constructing a cooling efficiency sequence according to the dynamic changes of the temperature sequence and the water volume sequence; Obtain building load demand data and ambient temperature data, and calculate the load demand change rate sequence; By using the cooling efficiency sequence and the load demand change rate sequence, a dynamic coupling relationship sequence is constructed to determine an energy transfer delay sequence; Full-cycle features are extracted from the dynamic coupling relationship sequence and the energy transfer delay sequence to generate an energy efficiency prediction sequence.

2. The energy efficiency evaluation and optimization method of the photovoltaic power generation cold storage system according to claim 1, characterized in that: The step of acquiring real-time power data and weather condition data of a photovoltaic power station and calculating a power fluctuation rate sequence includes: The photovoltaic output power and sunshine intensity are collected by sensors, and the power fluctuation rate is calculated by time series analysis method. If the power at the previous time point is zero and the current power is non-zero, the power fluctuation rate is set to the preset value; The power fluctuation rate and sunshine intensity are processed by the sliding window method, and the average power fluctuation rate and average sunshine intensity in continuous time periods are calculated. The Pearson correlation coefficient is used to analyze the correlation between the average power fluctuation rate and the average sunshine intensity, and the quantitative results of power fluctuation are obtained.

3. The energy efficiency evaluation and optimization method of the photovoltaic power generation cold storage system according to claim 2, characterized in that: The step of extracting characteristic values ​​from the power fluctuation rate sequence and determining a sequence of influence coefficients of fluctuations on ice-making energy input includes: The power fluctuation rate sequence is processed by the sliding window method to extract the smooth fluctuation feature sequence; The energy input per unit time is calculated by integration, and the instantaneous mapping relationship between the smooth fluctuation feature sequence and the energy input is analyzed using a convolutional neural network to obtain a mapping feature matrix. The influence coefficient sequence is calculated according to the mapping feature matrix. If the influence coefficient exceeds a preset threshold, the relationship between the smooth fluctuation feature sequence and the energy input is processed by the random forest algorithm, and the influence coefficient sequence is adjusted to obtain a dynamic influence sequence of fluctuation on energy input.

4. The energy efficiency evaluation and optimization method of the photovoltaic power generation cold storage system according to claim 1, characterized in that: The updating of the real-time energy input model through the influence coefficient sequence and the calculation of the dynamic changes of the temperature sequence and the water volume sequence include: The energy sequence is updated through the influence coefficient sequence and energy input calculation, and the temperature sequence and water volume sequence are generated by using the difference equation; The temperature series and water volume series are processed by the sliding window method to obtain smooth temperature series and smooth water volume series. The mapping relationship between the smoothed temperature series and the updated energy series is analyzed by linear regression to obtain the temperature-energy correlation coefficient matrix. The temperature energy correlation coefficient matrix and the water volume energy correlation coefficient matrix are combined through a matrix fusion method to generate a comprehensive dynamic influence matrix, and the temperature sequence and the water volume sequence are adjusted according to the comprehensive dynamic influence matrix.

5. The energy efficiency evaluation and optimization method of the photovoltaic power generation cold storage system according to claim 1, characterized in that: The step of constructing a cooling efficiency sequence according to the dynamic changes of the temperature sequence and the water volume sequence includes: The initial cooling efficiency sequence is calculated by a preset efficiency weight coefficient, and the initial cooling efficiency sequence and historical data are input into a support vector machine model to generate an efficiency prediction sequence; the efficiency weight coefficient is determined by a multivariate linear regression of historical cooling data, wherein the temperature weight γ ranges from 0.32 to 0.45, and the water weight ranges from 0.55 to 0.68; The efficiency prediction sequence is processed by the sliding window method to obtain a smooth efficiency sequence; The least square method is used to analyze the linear relationship between the smoothed efficiency series and the temperature series, and the temperature efficiency correlation coefficient matrix is ​​obtained; The temperature efficiency correlation coefficient matrix and the water efficiency correlation coefficient matrix are fused by matrix weighting to generate a comprehensive efficiency influence matrix, and the smoothed efficiency sequence is adjusted according to the comprehensive efficiency influence matrix.

6. The energy efficiency evaluation and optimization method of the photovoltaic power generation cold storage system according to claim 1, characterized in that: The step of obtaining building load demand data and ambient temperature data and calculating a load demand change rate sequence includes: Decomposing the load demand data through the long short-term memory network, obtaining the trend component and the period component, and calculating the load demand change rate sequence; The Pearson correlation coefficient is used to analyze the correlation between the load demand change rate sequence and the ambient temperature data, and the correlation coefficient sequence is obtained; The correlation coefficient sequence is processed by the sliding window method to obtain a smoothed correlation coefficient sequence; A random forest model is used to fit the nonlinear relationship between the smoothed correlation coefficient sequence and the ambient temperature data to generate a fitting relationship function, and the load demand change rate sequence is adjusted according to the fitting relationship function.

7. The method for optimizing energy efficiency evaluation of a photovoltaic power generation and cold storage system according to claim 1, characterized in that: The step of constructing a dynamic coupling relationship sequence and determining an energy transfer delay sequence through the cooling efficiency sequence and the load demand change rate sequence includes: The dynamic coupling relationship sequence is calculated by the preset coupling weight. If the dynamic coupling relationship sequence exceeds the preset threshold, the time lag effect characteristics are calculated by the sliding window method to generate an energy transfer delay sequence. The support vector machine algorithm is used to process the dynamic coupling relationship sequence and the energy transfer delay sequence to obtain the mapping relationship function; The mapping relationship function and the cooling efficiency sequence are analyzed through the long short-term memory network, the dynamic coupling relationship sequence in the next time period is predicted, and the dynamic coupling relationship sequence is adjusted according to the prediction results.

8. The energy efficiency evaluation and optimization method of the photovoltaic power generation and cold storage system according to claim 1, characterized in that: The step of extracting full-cycle features from the dynamic coupling relationship sequence and the energy transfer delay sequence to generate an energy efficiency prediction sequence includes: The periodic characteristics of the dynamic coupling relationship sequence and the energy transfer delay sequence are extracted through preprocessing methods, and the random forest algorithm is used to perform multivariate regression on the temperature sequence, water volume sequence, cooling efficiency sequence and load demand change rate sequence to generate a preliminary energy efficiency prediction sequence. The fluctuation characteristic sequence of the preliminary energy efficiency prediction sequence is calculated by the sliding window method. If the fluctuation characteristic sequence exceeds a preset threshold, the support vector machine algorithm is used to process the fluctuation characteristic sequence and the energy transfer delay sequence to obtain a mapping characteristic sequence, and the energy efficiency prediction sequence is adjusted according to the mapping characteristic sequence.

9. An energy efficiency evaluation and optimization system for a photovoltaic power generation and cold storage system for implementing any one of the methods described in claims 1 to 8, characterized in that: Included are: Data acquisition module, used to obtain real-time power data of photovoltaic power plants, weather condition data, ice storage tank temperature data, water volume data, building load demand data and ambient temperature data through sensor networks; The power fluctuation analysis module is used to calculate the power fluctuation rate sequence based on the sliding window algorithm and the Pearson correlation coefficient to generate the power fluctuation quantitative results; Dynamic modeling module, used to process smooth fluctuation feature sequences through convolutional neural networks and random forest algorithms, generate real-time energy input models, and output dynamic change parameters of temperature sequences and water volume sequences; The coupling relationship building module is used to integrate the support vector machine model and the long short-term memory network to process the time lag effect of the cooling efficiency sequence and the load demand change rate sequence, and generate a dynamic coupling relationship sequence and an energy transfer delay sequence; Energy efficiency prediction engine module, which is used to perform multivariate regression analysis and fluctuation feature mapping processing, and generate a full-cycle energy efficiency prediction sequence based on the comprehensive dynamic impact matrix and period feature extraction; The feedback optimization module is used to adaptively calibrate the influence coefficient sequence and dynamic coupling relationship sequence according to the preset threshold value, and update the system evaluation parameters through the matrix weighted fusion algorithm.

Images