Intelligent control method and device for photovoltaic energy storage power consumption in industrial parks

Abstract

The present application belongs to the field of power control technology, and discloses a method and device for intelligent control of photovoltaic energy storage electricity consumption in industrial parks. The method and device combine the historical operation data of the equipment, the real-time ambient temperature and the real-time humidity to predict the total reactive power demand of the power electronic equipment, and perform dynamic reactive power compensation based on this. At the same time, the photovoltaic inverter and the energy storage converter are coordinated to perform voltage stabilization control at the grid connection point, and the photovoltaic energy storage system cooling system is controlled according to the real-time ambient temperature and real-time humidity. The method and device can effectively adapt to high temperature and high humidity environments and effectively cope with dynamic changes in reactive power.

Description

Technical Field

[0001] The present application relates to the field of power control technology, and more specifically, to an intelligent control method and device for photovoltaic energy storage power consumption in industrial parks. Background Art

[0002] To reduce electricity costs, subtropical coastal industrial parks are widely deploying photovoltaic energy storage systems. However, the region's persistent high temperatures and high humidity in the summer pose a significant challenge to the stable operation of photovoltaic energy storage systems. High temperatures significantly reduce the power generation efficiency of photovoltaic modules, while high humidity exacerbates energy loss during the charging and discharging of storage batteries. Furthermore, the industrial inductive loads deployed within the industrial park and the power electronics in the photovoltaic energy storage system operate under the combined effects of high temperature and humidity, significantly increasing their reactive power demand. To ensure the reliability of the industrial park's power supply, intelligent control systems urgently need to achieve effective reactive power compensation and voltage stability control at the grid connection point.

[0003] However, traditional reactive power compensation methods, such as fixed capacitors or traditional reactive compensators, have been unable to effectively adapt to the rapid fluctuations and nonlinear changes in reactive power demand in the high temperature and high humidity environment of subtropical coastal industrial parks. Static compensation strategies cannot dynamically adjust the compensation amount according to real-time changes in ambient temperature and humidity, and are prone to under-compensation or over-compensation, which in turn seriously affects the stability of the grid voltage and the overall operating efficiency of the system. Furthermore, high temperature and high humidity environments will accelerate the performance degradation of power electronic equipment and energy storage equipment, further exacerbating the complexity and uncertainty of reactive power demand. Therefore, there is an urgent need for an intelligent control method that can fully adapt to high temperature and high humidity environments and effectively cope with dynamic changes in reactive power, so as to effectively ensure the stable operation and efficient use of electricity in photovoltaic energy storage systems in subtropical coastal industrial parks.

[0004] In view of the above problems, the existing technology is in urgent need of improvement. Summary of the Invention

[0005] The purpose of this application is to provide an intelligent control method and device for photovoltaic energy storage electricity consumption in industrial parks, which can effectively adapt to high temperature and high humidity environments and effectively cope with dynamic changes in reactive power.

[0006] In a first aspect, the present application provides an intelligent control method for photovoltaic energy storage power consumption in an industrial park, which is used to control power consumption of a photovoltaic energy storage system in a subtropical coastal industrial park. The method comprises:

[0007] S1. Obtain the real-time ambient temperature and humidity in the area where the PV energy storage system is deployed;

[0008] S2. Use a combined prediction model to predict the total reactive power demand of power electronic equipment in the industrial park and the photovoltaic energy storage system based on historical equipment operating data, real-time ambient temperature, and real-time humidity.

[0009] S3. According to the prediction results of the combined prediction model, dynamically adjust the output of the reactive compensation device to achieve reactive power compensation;

[0010] S4. Adjust the operating modes of the photovoltaic inverter and energy storage converter to collaboratively control the output of active and reactive power to maintain voltage stability at the grid connection point;

[0011] S5. Optimize the operation strategy of the photovoltaic energy storage system cooling system based on the real-time ambient temperature and real-time humidity to reduce the equipment operating temperature.

[0012] This method combines historical equipment operation data, real-time ambient temperature and real-time humidity to predict the total reactive power demand of power electronic equipment, and performs dynamic reactive power compensation based on this. At the same time, it coordinates photovoltaic inverters and energy storage converters to stabilize the voltage at the grid connection point, and controls the cooling system of the photovoltaic energy storage system according to the real-time ambient temperature and real-time humidity. It can effectively adapt to high temperature and high humidity environments and effectively cope with dynamic changes in reactive power.

[0013] Preferably, the combined prediction model includes a linear regression model and a nonlinear neural network model; the input of the linear regression model includes the active power, ambient temperature and humidity in the historical operation data of the equipment, and the output is a first reactive power prediction value; the input of the nonlinear neural network model includes the active power, ambient temperature and humidity in the historical operation data, and the output is a second reactive power prediction value;

[0014] Step S2 includes:

[0015] S201. Based on the active power, ambient temperature and humidity in the historical operation data of the device, a linear regression model is used to obtain a first reactive power prediction value, and a nonlinear neural network model is used to obtain a second reactive power prediction value;

[0016] S202. Calculate a first weight of the linear regression model and a second weight of the nonlinear neural network model using a dynamic weight allocation method based on historical prediction errors of the linear regression model and the nonlinear neural network model, where the sum of the first weight and the second weight is 1;

[0017] S203. An initial reactive power prediction value is obtained by weighted average calculation based on the first reactive power prediction value, the second reactive power prediction value, the first weight, and the second weight;

[0018] S204. Based on the real-time ambient temperature and real-time humidity, a Sigmoid function is used to perform nonlinear correction on the initial reactive power prediction value to obtain a final reactive power prediction value.

[0019] Through the above method, the prediction accuracy of reactive power can be effectively improved.

[0020] Preferably, step S202 includes:

[0021] A1. Perform sliding window processing on historical forecast errors to construct an error sequence. Then, use a wavelet decomposition algorithm to decompose the error sequence to obtain multiple subsequences of different frequencies.

[0022] A2. For each subsequence, calculate the autocorrelation coefficient. Based on the autocorrelation coefficient, use a gradient descent algorithm with an adaptive step size to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. The sum of the first and second weights is 1.

[0023] Therefore, the step size of weight adjustment can be dynamically adjusted according to the stability of the error, thereby improving the efficiency and accuracy of weight adjustment.

[0024] Preferably, step A2 includes:

[0025] B1. For each subsequence, calculate the autocorrelation coefficient. If the autocorrelation coefficient is greater than a preset threshold, the corresponding subsequence is determined to be a stationary sequence and step B2 is executed. Otherwise, the corresponding subsequence is determined to be a non-stationary sequence and step B3 is executed.

[0026] B2. For the stationary sequence, use a fixed-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. These weights are denoted as the first preferred weight value and the second preferred weight value, respectively. The fixed step size is the preset step size.

[0027] B3. For non-stationary sequences, use a variable-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. These are denoted as the first preferred weight value and the second preferred weight value, respectively. The variable-step calculation formula is: st_i = st_max * exp(-|r_i|), where st_i is the step size of the i-th subsequence, st_max is the preset maximum step size, and r_i is the autocorrelation coefficient of the i-th subsequence.

[0028] B4. Perform a weighted average of the first weight preference value and the second weight preference value calculated for all subsequences to obtain the final first weight of the linear regression model and the final second weight of the nonlinear neural network model. The sum of the final first weight and the final second weight is 1.

[0029] Preferably, step S3 includes:

[0030] S301 real-time collection of power consumption data of various types of loads in the industrial park, identifying the first type of load that responds quickly to reactive power demand and the second type of load that responds slowly;

[0031] S302. Based on the historical power consumption data ratio of the first and second types of loads, the total reactive power demand is decomposed into the reactive power demand forecast value of the first type of load and the reactive power demand forecast value of the second type of load;

[0032] S303. For the reactive power demand forecast value of the first type of load, a feedforward control method is used to directly adjust the output of the reactive compensation device according to the forecast value to achieve rapid compensation;

[0033] S304. Based on the reactive power demand forecast value of the second type of load, a feedback control method is adopted to monitor the grid voltage in real time. According to the grid voltage deviation and the reactive power demand forecast value of the second type of load, the output of the reactive compensation device is adjusted through the PID controller to achieve accurate compensation.

[0034] Preferably, step S301 includes:

[0035] Real-time collection of electricity consumption data of various types of loads in the industrial park to build an electricity consumption data set; electricity consumption data includes active power, reactive power, voltage, and current;

[0036] Preprocess the electricity consumption dataset to obtain a preprocessed electricity consumption dataset; the preprocessing includes missing value filling, outlier detection and elimination, and sliding average filtering;

[0037] Extract the time domain features and frequency domain features of the preprocessed electricity consumption dataset. The time domain features include mean, variance, peak value and kurtosis, and the frequency domain features include total harmonic distortion rate to obtain the load feature vector.

[0038] The K-means clustering algorithm is used to divide the load into the first type of load and the second type of load according to the load feature vector.

[0039] Preferably, step S4 includes:

[0040] S401 real-time monitoring of the grid connection point voltage, calculate the grid connection point voltage and the preset voltage deviation, and obtain the current active power output value of the photovoltaic inverter and energy storage converter;

[0041] S402. Based on the voltage deviation and the current active power output value, a droop control algorithm is used to calculate the reactive power regulation instruction of the photovoltaic inverter and the energy storage converter, wherein the droop coefficient in the droop control algorithm is adjusted according to the real-time ambient temperature and real-time humidity;

[0042] S403. According to the reactive power regulation instruction, adjust the working mode of the photovoltaic inverter and the energy storage converter, and coordinate the output of active power and reactive power. Among them, the photovoltaic inverter prioritizes the adjustment of reactive power. When the reactive power of the photovoltaic inverter reaches the upper limit, the energy storage converter continues to adjust the reactive power;

[0043] S404. Monitor the operating status of the photovoltaic inverter and energy storage converter in real time. If an equipment overload or failure is detected, start the backup inverter or energy storage converter and adjust the operating parameters to ensure stable voltage at the grid connection point.

[0044] Preferably, step S402 includes:

[0045] According to the real-time ambient temperature and real-time humidity, the temperature correction coefficient and humidity correction coefficient are calculated by fuzzy algorithm;

[0046] Calculate the comprehensive environmental correction factor based on the temperature correction factor and the humidity correction factor;

[0047] Adjust the droop coefficient of the droop control algorithm according to the comprehensive environmental correction coefficient;

[0048] According to the voltage deviation and the current active power output value, the adjusted droop coefficient is used to calculate the reactive power regulation instructions of the photovoltaic inverter and the energy storage converter based on the droop control algorithm.

[0049] Preferably, step S5 includes:

[0050] S501. Real-time monitoring of the inverter temperature, energy storage battery temperature, and converter temperature of the photovoltaic energy storage system to build a device temperature data set;

[0051] S502. Based on the real-time ambient temperature, real-time humidity and device temperature data set, a fuzzy control algorithm is used to calculate the cooling demand level;

[0052] S503. According to the cooling demand level and the current operating state of the cooling system, a hierarchical control strategy is used to adjust the operating parameters of the cooling system.

[0053] In a second aspect, the present application provides an intelligent control device for photovoltaic energy storage power consumption in an industrial park, which is used to control power consumption of a photovoltaic energy storage system in a subtropical coastal industrial park. The device includes:

[0054] Environmental parameter monitoring module, used to obtain the real-time ambient temperature and humidity in the area where the photovoltaic energy storage system is deployed;

[0055] The reactive power demand forecasting module is used to predict the total reactive power demand of power electronic equipment in industrial parks and photovoltaic energy storage systems using a combined forecasting model based on historical equipment operating data, real-time ambient temperature, and real-time humidity;

[0056] Dynamic reactive power compensation control module, used to dynamically adjust the output of reactive power compensation device according to the prediction results of combined prediction model to achieve reactive power compensation;

[0057] The voltage stabilization control module is used to adjust the operating mode of the photovoltaic inverter and energy storage converter, coordinate the output of active power and reactive power, and maintain voltage stability at the grid connection point;

[0058] The equipment cooling optimization module is used to optimize the operation strategy of the photovoltaic energy storage system cooling system according to the real-time ambient temperature and real-time humidity, and reduce the equipment operating temperature.

[0059] Beneficial effects: The present application provides an intelligent control method and device for photovoltaic energy storage electricity consumption in industrial parks. The method and device combine the historical operation data of the equipment, the real-time ambient temperature and the real-time humidity to predict the total reactive power demand of the power electronic equipment, and perform dynamic reactive power compensation based on this. At the same time, the photovoltaic inverter and the energy storage converter are coordinated to perform voltage stability control at the grid connection point, and the photovoltaic energy storage system cooling system is controlled according to the real-time ambient temperature and real-time humidity. The method and device can effectively adapt to high temperature and high humidity environments and effectively respond to dynamic changes in reactive power. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Flowchart of the intelligent control method for photovoltaic energy storage electricity consumption in an industrial park provided in an embodiment of the present application.

[0061] Figure 2 This is a schematic diagram of the structure of the intelligent control device for photovoltaic energy storage and electricity consumption in an industrial park provided in an embodiment of the present application.

[0062] Explanation of reference numbers: 1. Environmental parameter monitoring module; 2. Reactive power demand prediction module; 3. Dynamic reactive power compensation control module; 4. Voltage stabilization control module; 5. Equipment cooling optimization module. DETAILED DESCRIPTION

[0063] The technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for which protection is claimed, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0064] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0065] refer to Figure 1 This application proposes an intelligent control method for photovoltaic energy storage power consumption in industrial parks, which is used to control the power consumption of photovoltaic energy storage systems in subtropical coastal industrial parks. The method includes:

[0066] S1. Obtain the real-time ambient temperature and humidity in the area where the PV energy storage system is deployed;

[0067] S2. Use a combined prediction model to predict the total reactive power demand of power electronic equipment in the industrial park and the photovoltaic energy storage system based on historical equipment operating data, real-time ambient temperature, and real-time humidity.

[0068] S3. According to the prediction results of the combined prediction model, dynamically adjust the output of the reactive compensation device to achieve reactive power compensation;

[0069] S4. Adjust the operating modes of the photovoltaic inverter and energy storage converter to collaboratively control the output of active and reactive power to maintain voltage stability at the grid connection point;

[0070] S5. Optimize the operation strategy of the photovoltaic energy storage system cooling system based on the real-time ambient temperature and real-time humidity to reduce the equipment operating temperature.

[0071] In step S1, ambient temperature sensors and humidity sensors deployed in the photovoltaic energy storage system area can be used to monitor and collect ambient temperature and humidity data in real time, which can be used as basic data for subsequent reactive power prediction and control.

[0072] In step S2, the combined prediction model can integrate a linear regression model and a nonlinear neural network model. The linear regression model analyzes the linear relationship between the load's historical active power data and the ambient temperature and humidity to provide a preliminary reactive power forecast. The nonlinear neural network model captures the complex nonlinear relationship between ambient temperature and humidity and the total reactive power demand, further improving prediction accuracy. The prediction results of the two models can be combined using a dynamic weight allocation method to achieve a more accurate total reactive power demand forecast.

[0073] In step S3, the reactive power compensation device is driven according to the predicted value to dynamically adjust the compensation amount. The output of the reactive power compensation device can quickly respond to changes in reactive power demand to achieve dynamic compensation.

[0074] In step S4, the voltage at the grid connection point is monitored in real time. If voltage deviations occur, the operating modes of the PV inverter and energy storage converter are adjusted simultaneously. The PV inverter and energy storage converter work together to maintain the voltage at the grid connection point within a set range by adjusting the output of active and reactive power.

[0075] In step S5, a fuzzy control algorithm can be used to analyze the relationship between environmental parameters and device temperature to calculate the cooling requirement level. The cooling system, such as a fan or liquid cooling system, adjusts operating parameters, such as fan speed or coolant flow, based on the cooling requirement level to optimize cooling efficiency and reduce device operating temperature.

[0076] Specifically, this application provides an intelligent control method for photovoltaic energy storage power consumption in industrial parks, addressing the technical challenges of reactive power compensation and voltage stability in photovoltaic energy storage systems in subtropical coastal industrial parks under varying ambient temperature and humidity. This method first obtains the real-time ambient temperature and humidity in the area where the photovoltaic energy storage system is deployed, providing environmental parameter references for subsequent control. Then, based on historical equipment operating data, real-time ambient temperature, and real-time humidity, a combined prediction model is used to predict the total reactive power demand of the power electronics equipment in the industrial park and the photovoltaic energy storage system. This combined prediction model considers the impact of ambient temperature and humidity on total reactive power demand, improving prediction accuracy. Subsequently, based on the prediction results of the combined prediction model, the output of the reactive power compensation device is dynamically adjusted to achieve reactive power compensation. Dynamic adjustment compensates based on predicted demand, avoiding the under- or over-compensation issues of traditional static compensation strategies. Furthermore, the operating modes of the photovoltaic inverter and energy storage converter are adjusted to coordinately control the output of active and reactive power to maintain voltage stability at the grid connection point. This coordinated control ensures grid voltage stability and guarantees power supply reliability. Finally, based on the real-time ambient temperature and humidity, the operating strategy of the photovoltaic energy storage system's cooling system is optimized to reduce equipment operating temperature. Optimizing cooling system operation strategies reduces equipment temperature, improving system operating efficiency and equipment lifespan. The proposed method can dynamically adjust reactive power compensation and voltage control strategies based on the high temperature and high humidity environments unique to subtropical coastal industrial parks, while simultaneously optimizing equipment cooling to ensure stable operation and efficient power consumption of photovoltaic energy storage systems.

[0077] In some possible implementations, the combined prediction model includes a linear regression model and a nonlinear neural network model; the linear regression model input includes active power, ambient temperature, and humidity in the historical operation data of the device, and the output is a first reactive power prediction value; the nonlinear neural network model input includes active power, ambient temperature, and humidity in the historical operation data, and the output is a second reactive power prediction value;

[0078] Step S2 includes:

[0079] S201. Based on the active power, ambient temperature and humidity in the historical operation data of the device, a linear regression model is used to obtain a first reactive power prediction value, and a nonlinear neural network model is used to obtain a second reactive power prediction value;

[0080] S202. Calculate a first weight of the linear regression model and a second weight of the nonlinear neural network model using a dynamic weight allocation method based on historical prediction errors of the linear regression model and the nonlinear neural network model, where the sum of the first weight and the second weight is 1;

[0081] S203. An initial reactive power prediction value is obtained by weighted average calculation based on the first reactive power prediction value, the second reactive power prediction value, the first weight, and the second weight;

[0082] S204. Based on the real-time ambient temperature and real-time humidity, a Sigmoid function is used to perform nonlinear correction on the initial reactive power prediction value to obtain a final reactive power prediction value.

[0083] Among them, step S201 refers to using two different types of prediction models, namely linear regression models and nonlinear neural network models, to make preliminary predictions of reactive power. Specifically, a variety of linear regression models can be used, such as simple linear regression and multivariate linear regression, to capture the linear relationship between the total reactive power demand and active power, ambient temperature, and humidity. The nonlinear neural network model can use a backpropagation neural network, a convolutional neural network, or a recurrent neural network, etc., to learn and simulate the complex nonlinear relationship between the total reactive power demand and input parameters. The historical operation data of the equipment includes but is not limited to historical active power data, historical ambient temperature data, and historical humidity data (the active power, ambient temperature, and humidity in the historical operation data of the equipment, namely historical active power data, historical ambient temperature data, and historical humidity data).

[0084] Among them, step S202 refers to dynamically allocating weights based on the historical prediction errors of the two models. Specifically, the historical prediction error can be obtained by calculating the difference between the model prediction value and the actual value. The dynamic weight allocation method can be implemented using a variety of algorithms, such as gradient descent, genetic algorithm, or particle swarm optimization algorithm. The principle of weight allocation is that the model with smaller error is assigned a higher weight, and the model with larger error is assigned a lower weight. The limitation that the sum of the weights is 1 ensures that the weight allocation ratio of the two models is reasonable and easy to calculate.

[0085] Step S203 involves taking a weighted average of the prediction results from the two models. Specifically, by multiplying the first reactive power prediction value by a first weight and the second reactive power prediction value by a second weight, and adding the two together, an initial reactive power prediction value is obtained that comprehensively considers both linear and nonlinear relationships. This weighted average approach leverages the advantages of both models and reduces the risk of single-model predictions.

[0086] Among them, step S204 refers to using the Sigmoid function to perform nonlinear correction on the initial reactive power prediction value. Specifically, due to its S-shaped characteristics, the Sigmoid function can map the input value to a range between 0 and 1, or other preset nonlinear ranges, to achieve nonlinear adjustment of the initial prediction value. Real-time ambient temperature and real-time humidity are used as inputs of the Sigmoid function, so that the correction process can reflect changes in environmental factors in real time. The specific form of the Sigmoid function can be selected and adjusted according to the actual application scenario, such as the standard Sigmoid function, the hyperbolic tangent Sigmoid function, etc.

[0087] Specifically, for the problem of forecasting total reactive power demand in industrial parks, considering that total reactive power demand is influenced by multiple factors and exhibits a mixture of linear and nonlinear characteristics, a combined forecasting model was designed that includes a linear regression model and a nonlinear neural network model. The linear regression model effectively captures the linear relationship between total reactive power demand, active power, and ambient temperature and humidity, while the nonlinear neural network model excels at learning and fitting complex nonlinear relationships. Combining these two models allows for more comprehensive extraction of information from the data, improving the adaptability and accuracy of the forecasting model. A dynamic weight allocation method was introduced as a model fusion strategy. This method adaptively adjusts the weight of each model in the combined forecast based on its historical forecast performance. In specific implementation, a sliding window mechanism can be used to regularly evaluate the error of each model in recent forecasts. Models with smaller errors are given higher weights, while those with smaller errors are given lower weights. This allows the combined forecasting model to dynamically track the forecasting performance of each model, achieving a more optimal combined forecasting result. To further improve forecasting accuracy, especially when ambient temperature and humidity have nonlinear effects on total reactive power demand, a sigmoid function is introduced to correct the forecast results. The Sigmoid function, with its S-shaped curve characteristics, maps input values ​​to a range between 0 and 1, introducing nonlinear transformations. By using real-time ambient temperature and humidity as inputs to the Sigmoid function, nonlinear adjustments can be made to the initial predicted values, making the final prediction more consistent with reality. For example, in high-temperature and high-humidity environments, total reactive power demand may show an accelerating growth trend. The Sigmoid function can simulate this nonlinear growth characteristic and adjust the predicted value upward.

[0088] In some preferred embodiments, step S202 includes:

[0089] A1. Perform sliding window processing on historical forecast errors to construct an error sequence. Then, use a wavelet decomposition algorithm to decompose the error sequence to obtain multiple subsequences of different frequencies.

[0090] A2. For each subsequence, calculate the autocorrelation coefficient. Based on the autocorrelation coefficient, use a gradient descent algorithm with an adaptive step size to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. The sum of the first and second weights is 1.

[0091] Among them, step A1 and step A2 are specific limitations of the dynamic weight allocation method. The dynamic weight allocation method aims to adaptively adjust the weights of the linear regression model and the nonlinear neural network model in the combined prediction based on the performance of the model in historical predictions, so that the prediction model can better adapt to changing needs. However, a simple dynamic weight allocation method may not be able to fully cope with the complexity of the prediction error. The error may contain multiple frequency components. For example, some error components change rapidly and have a higher frequency, while other error components change slowly and have a lower frequency. In order to handle these complex errors more finely, step S202 introduces wavelet decomposition technology.

[0092] Step A1 preprocesses historical forecast errors. Specifically, a sliding window is first used to process historical forecast errors and construct an error sequence. The size of the sliding window can be adjusted according to actual needs. For example, it can be selected to include error data from the most recent period to capture recent error trends. The purpose of sliding window processing is to organize discrete historical forecast errors into a continuous sequence, facilitating subsequent wavelet decomposition. The error sequence is then decomposed using a wavelet decomposition algorithm to obtain multiple subsequences of different frequencies. Wavelet decomposition is a signal processing technique that decomposes a complex signal into multiple sub-signals of different frequencies, each representing a component of the original signal within a specific frequency range. For a forecast error sequence, wavelet decomposition can decompose the error into components of different frequencies. For example, high-frequency components may reflect random noise or sudden disturbances, while low-frequency components may reflect systematic deviations or trend-related errors. There are various specific implementation algorithms for wavelet decomposition, such as Haar wavelets, Daubechies wavelets, and Symlets wavelets. The appropriate wavelet basis function and number of decomposition levels can be selected based on the actual application scenario and the characteristics of the error sequence. Through wavelet decomposition, the complex error signal can be decomposed into multiple relatively simple subsequences, laying the foundation for subsequent differentiated processing of different frequency error components.

[0093] Step A2 calculates weights for each frequency subsequence. Specifically, for each subsequence obtained in step A1, the autocorrelation coefficient is calculated. The autocorrelation coefficient is an indicator that measures the degree of correlation within a sequence and can reflect the correlation between values ​​at different time points in the sequence. For a prediction error subsequence, the autocorrelation coefficient can reflect the temporal dependence or regularity of the errors in that frequency component. A high autocorrelation coefficient for a subsequence indicates that the errors at that frequency have strong regularity and predictability; conversely, a low autocorrelation coefficient indicates that the errors at that frequency are highly random. There are various methods for calculating the autocorrelation coefficient, for example, the Pearson correlation coefficient or the Spearman rank correlation coefficient can be used. Based on the calculated autocorrelation coefficient, a gradient descent algorithm with an adaptive step size is used to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. The gradient descent algorithm is a commonly used optimization algorithm that can be used to find the minimum value of a function. In this solution, the goal of the gradient descent algorithm is to minimize the prediction error of the combined prediction model. By continuously iteratively adjusting the weights of the linear regression model and the nonlinear neural network model, the prediction results of the combined prediction model are closer to the true value. Adaptive step size means that during the iteration process of the gradient descent algorithm, the step size can be dynamically adjusted according to the characteristics of the error. For example, when the error has a strong autocorrelation, the step size can be appropriately increased to accelerate the convergence speed; when the error autocorrelation is weak, the step size can be appropriately reduced to avoid oscillation or divergence. There are many specific implementation methods for adaptive step size. For example, the step size can be dynamically adjusted according to factors such as the size of the autocorrelation coefficient, the size of the gradient, or the number of iterations. Through the gradient descent algorithm with adaptive step size, the weights of the linear regression model and the nonlinear neural network model can be calculated more efficiently and accurately, so that the weight distribution can better adapt to the characteristics of different frequency error components. The sum of the first weight and the second weight is limited to 1, which ensures that the weight distribution of the two models constitutes a complete weight space and avoids redundancy or conflict in weight distribution.

[0094] Specifically, before dynamic weight allocation, this solution first performs wavelet decomposition on the historical forecast error, breaking it down into multiple subsequences of different frequencies. This approach has the advantage of decomposing complex error signals into different frequency domains, enabling more refined analysis and processing of error components at different frequencies. For each frequency subsequence, the autocorrelation coefficient is calculated to assess the predictability of the frequency error. A higher autocorrelation coefficient indicates greater regularity in the frequency error, while a lower one indicates greater randomness. The step size of the gradient descent algorithm is then adaptively adjusted based on the autocorrelation coefficient to calculate the weights for the linear regression model and the nonlinear neural network model. For frequency subsequences with higher autocorrelation coefficients, a larger step size is used to accelerate weight adjustment; for frequency subsequences with lower autocorrelation coefficients, a smaller step size is used to avoid over-adjustment. This approach allows weight allocation to more precisely adapt to the characteristics of different frequency errors, thereby improving the prediction accuracy of the combined forecasting model.

[0095] In some embodiments, step A2 comprises:

[0096] B1. For each subsequence, calculate the autocorrelation coefficient. If the autocorrelation coefficient is greater than a preset threshold, the corresponding subsequence is determined to be a stationary sequence and step B2 is executed. Otherwise, the corresponding subsequence is determined to be a non-stationary sequence and step B3 is executed.

[0097] B2. For the stationary sequence, use a fixed-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. These weights are denoted as the first preferred weight value and the second preferred weight value, respectively. The fixed step size is the preset step size.

[0098] B3. For non-stationary sequences, use a variable-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. These are denoted as the first preferred weight value and the second preferred weight value, respectively. The variable-step calculation formula is: st_i = st_max * exp(-|r_i|), where st_i is the step size of the i-th subsequence, st_max is the preset maximum step size, and r_i is the autocorrelation coefficient of the i-th subsequence.

[0099] B4. Perform a weighted average of the first weight preference value and the second weight preference value calculated for all subsequences to obtain the final first weight of the linear regression model and the final second weight of the nonlinear neural network model. The sum of the final first weight and the final second weight is 1.

[0100] To address the differences in characteristics between subsequences of different frequencies within the error sequence, a refined weight allocation strategy was implemented. Specifically, the autocorrelation coefficient, which characterizes the stationarity of each subsequence obtained through wavelet decomposition, was calculated. A preset threshold was used to distinguish between stationary and non-stationary subsequences and could be set as needed. For example, the threshold could be set to 0.5. Subsequences with an autocorrelation coefficient greater than 0.5 were considered stationary; otherwise, they were considered non-stationary. This effectively identified the internal characteristics of the error sequence.

[0101] Furthermore, differentiated weight calculation methods are employed for different subsequence types. For stationary subsequences, due to their low volatility and relatively stable prediction errors, a fixed-step-size gradient descent algorithm is used to ensure the stability of weight updates. The preset step size can be adjusted based on the actual application scenario; for example, it can be set to 0.01. For non-stationary subsequences, due to their high data volatility and unstable prediction errors, a variable-step-size gradient descent algorithm is employed. The variable-step-size calculation formula is: st_i = st_max * exp(-|r_i|). Here, st_max is the preset maximum step size, which can be set as needed, for example, to 0.1, and r_i is the autocorrelation coefficient of the i-th subsequence. Therefore, the step size is negatively correlated with the autocorrelation coefficient. That is, the lower the autocorrelation of the error sequence (the greater the non-stationarity), the larger the step size. This allows the weights to quickly adapt to changes in the error, improving the model's flexibility and tracking capabilities.

[0102] Finally, by taking a weighted average of the first and second preferred weights calculated for all subsequences, we arrive at the final first weights for the linear regression model and the final second weights for the nonlinear neural network model. This weighted average comprehensively considers the weight calculation results for each subsequence to arrive at the optimal overall weight distribution.

[0103] Specifically, this approach enables more refined adjustment of the weights of the linear regression model and the nonlinear neural network model, fully accounting for the complex characteristics of the error sequence. This overcomes the issue of insufficient precision in weight allocation and improves the prediction accuracy of the combined prediction model. This improves the accuracy and control of reactive power compensation, ensuring stable operation and efficient power consumption of photovoltaic energy storage systems in subtropical coastal industrial parks.

[0104] For example, in one specific embodiment, the preset threshold is set to 0.6, the preset step size is set to 0.005, and the maximum step size st_max is set to 0.05. For the four error subsequences obtained by wavelet decomposition, the autocorrelation coefficients r_1, r_2, r_3, and r_4 are calculated respectively, assuming r_1 = 0.7, r_2 = 0.8, r_3 = 0.4, and r_4 = 0.5. Since r_1 > 0.6 and r_2 > 0.6, the first and second subsequences are determined to be stationary sequences, and the optimal weight values ​​are calculated using a gradient descent algorithm with a fixed step size of 0.005. Since r_3 < 0.6 and r_4 < 0.6, the third and fourth subsequences are determined to be non-stationary sequences, and the weights are calculated using a gradient descent algorithm with a variable step size, with step sizes st_3 = 0.05*exp(-|0.4|) ≈ 0.0335 and st_4 = 0.05*exp(-|0.5|) ≈ 0.0303, respectively. Using the gradient descent algorithm, we iteratively calculate the first and second preferred weights for each subsequence: (w1_1, w1_2), (w2_1, w2_2), (w3_1, w3_2), and (w4_1, w4_2). The final first weight W1 is (w1_1 + w2_1 + w3_1 + w4_1) / 4, and the final second weight W2 is (w1_2 + w2_2 + w3_2 + w4_2) / 4.

[0105] In some preferred embodiments, step S3 includes:

[0106] S301 real-time collection of power consumption data of various types of loads in the industrial park, identifying the first type of load that responds quickly to reactive power demand and the second type of load that responds slowly;

[0107] S302. Based on the historical power consumption data ratio of the first and second types of loads, the total reactive power demand is decomposed into the reactive power demand forecast value of the first type of load and the reactive power demand forecast value of the second type of load;

[0108] S303. For the reactive power demand forecast value of the first type of load, a feedforward control method is used to directly adjust the output of the reactive compensation device according to the forecast value (referring to the reactive power demand forecast value of the first type of load) to achieve rapid compensation;

[0109] S304. Based on the reactive power demand forecast value of the second type of load, a feedback control method is adopted to monitor the grid voltage in real time. According to the grid voltage deviation and the reactive power demand forecast value of the second type of load, the output of the reactive compensation device is adjusted through the PID controller to achieve accurate compensation.

[0110] In step S301, power usage data is collected in real time from various types of loads within the industrial park to comprehensively reflect the park's power usage characteristics. Load type identification can be achieved using a clustering algorithm, such as K-means clustering. This algorithm analyzes the historical power usage characteristics of the loads and divides them into first-class loads (fast response) and second-class loads (slow response). Fast and slow response speeds are mutually exclusive. Here, the fast-responding first-class load and the slow-responding second-class load refer to the first-class load responding faster than the second-class load.

[0111] In step S302, the total reactive power demand forecast value is provided by the combined forecast model. This forecast value is then decomposed based on the historical power usage data percentages of the two load types. This percentage data can be statistically obtained from the historical operation data of the equipment. For example, if the historical power usage data of the first load type accounts for 1 / 3 of the total historical power usage data, and the historical power usage data of the second load type accounts for 2 / 3 of the total historical power usage data, then the reactive power demand forecast value of the first load type is 1 / 3 of the total reactive power demand, and the reactive power demand forecast value of the second load type is 2 / 3 of the total reactive power demand.

[0112] In step S303 , for the feedforward control of the first type of load, the output of the reactive compensation device is directly adjusted according to the predicted value, and the control signal acts on the compensation device quickly.

[0113] In step S304, for feedback control of the second type of load, the PID controller is configured to adjust the output of the compensation device according to the grid voltage deviation and the predicted value, and the voltage deviation is used as a feedback signal by the PID controller to ensure compensation accuracy.

[0114] Specifically, to address the diverse response characteristics of loads within industrial parks, this method proposes a categorized dynamic reactive power compensation control strategy. First, in step S301, real-time electricity data collection and a clustering algorithm can be used to effectively categorize loads within the industrial park into two types with varying response speeds to reactive power demand. This allows the control strategy to be optimized for the characteristics of different load types. Second, in step S302, the total reactive power demand forecast is decomposed into the demand forecasts for each of the two load types, enabling refined allocation of the total reactive power demand. Furthermore, steps S303 and S304 employ differentiated control methods. For the first type of loads, which respond quickly, feedforward control is used to achieve rapid reactive power compensation, ensuring rapid responsiveness. For the second type of loads, which respond slowly, feedback control combined with a PID controller achieves precise reactive power compensation, ensuring compensation accuracy. Through the above steps, this method can address the reactive power compensation requirements of both fast-responding and slow-responding loads, improving overall reactive power compensation effectiveness and grid voltage stability.

[0115] In some possible implementations, step S301 includes:

[0116] Real-time collection of electricity consumption data of various types of loads in the industrial park to build an electricity consumption data set; electricity consumption data includes active power, reactive power, voltage, and current;

[0117] Preprocess the electricity consumption dataset to obtain a preprocessed electricity consumption dataset; the preprocessing includes missing value filling, outlier detection and elimination, and sliding average filtering;

[0118] Extract the time domain features and frequency domain features of the preprocessed electricity consumption dataset; the time domain features include mean, variance, peak value and kurtosis, and the frequency domain features include total harmonic distortion rate, to obtain the load feature vector;

[0119] The K-means clustering algorithm is used to divide the load into the first type of load and the second type of load according to the load feature vector.

[0120] To ensure classification accuracy, real-time electricity consumption data from various load types within the industrial park is first collected to form an electricity consumption dataset, laying the foundation for subsequent analysis. This electricity consumption data includes information such as active power, reactive power, voltage, and current, providing comprehensive insight into the load's operating status. Second, the collected electricity consumption data is preprocessed to produce a preprocessed electricity consumption dataset. This preprocessing process involves multiple steps. Missing value filling ensures data integrity. For example, linear interpolation is used to fill in small amounts of missing data that may occur during data collection. Outlier detection and elimination prevents abnormal data from interfering with analysis results. For example, an outlier detection method based on the 3σ principle is used to identify and eliminate data that deviates from the normal range. A sliding average filter is used to reduce the impact of data noise and improve data quality. For example, a sliding average filter with a sliding window size of 5 is used to smooth data fluctuations. Then, time and frequency domain features are extracted from the preprocessed data. Time domain features include mean, variance, peak, and kurtosis, while frequency domain features include total harmonic distortion. These features effectively characterize the load's electricity consumption characteristics. For example, the mean and variance reflect the average level and fluctuation of load power, the peak and kurtosis reflect the extreme values ​​and distribution of load power, and the total harmonic distortion rate reflects the degree of waveform distortion in the load current. Finally, a K-means clustering algorithm is used, using the load feature vector as input, to automatically classify the loads into the first and second categories. The K-means algorithm is an unsupervised learning algorithm that clusters data based on similarities, eliminating the need for manual intervention and enabling automatic load classification. For example, the K-means algorithm was set to have 2 cluster centers, Euclidean distance as the distance calculation method, and 100 iterations. Through these steps, accurate classification of loads within the industrial park is achieved, paving the way for the subsequent implementation of differentiated reactive power compensation control strategies for different load types. Data preprocessing ensures data quality and reduces noise interference. Feature extraction effectively characterizes load power characteristics. The application of the K-means clustering algorithm enables automatic load classification, improving classification efficiency and accuracy.

[0121] In some specific implementations, in order to further improve the accuracy and adaptability of load classification, more advanced data cleaning methods can be introduced in the preprocessing stage. For example, a denoising method based on wavelet analysis can more effectively remove high-frequency noise in electricity consumption data. In the feature extraction stage, more dimensions and more characterization features can be added. For example, frequency domain features can increase the content of each harmonic, and time domain features can increase waveform factors, margin factors, etc., to more comprehensively describe the power consumption characteristics of the load. In terms of clustering algorithms, DBSCAN clustering algorithms or hierarchical clustering algorithms can be tried to adapt to the load classification needs of industrial parks of different types and sizes. Through the optimization of preprocessing methods, enhancement of feature extraction, and improvement of clustering algorithms, the accuracy and robustness of load classification can be further improved, providing more reliable guarantees for the intelligent control of photovoltaic energy storage systems in industrial parks.

[0122] This application further proposes that step S4 includes:

[0123] S401 real-time monitoring of the grid connection point voltage, calculate the grid connection point voltage and the preset voltage deviation, and obtain the current active power output value of the photovoltaic inverter and energy storage converter;

[0124] S402. Based on the voltage deviation and the current active power output value, a droop control algorithm is used to calculate the reactive power regulation instruction of the photovoltaic inverter and the energy storage converter, wherein the droop coefficient in the droop control algorithm is adjusted according to the real-time ambient temperature and real-time humidity;

[0125] S403. According to the reactive power regulation instruction, adjust the working mode of the photovoltaic inverter and the energy storage converter, and coordinate the output of active power and reactive power. Among them, the photovoltaic inverter prioritizes the adjustment of reactive power. When the reactive power of the photovoltaic inverter reaches the upper limit, the energy storage converter continues to adjust the reactive power;

[0126] S404. Monitor the operating status of the photovoltaic inverter and energy storage converter in real time. If an equipment overload or failure is detected, start the backup inverter or energy storage converter and adjust the operating parameters to ensure stable voltage at the grid connection point.

[0127] In step S401, the grid connection point voltage can be obtained in real time through voltage sensor monitoring. The voltage deviation is the difference between the grid connection point voltage and a preset voltage, which is a target voltage value set according to grid operating standards. The current active power output values ​​of the photovoltaic inverter and energy storage converter can be obtained in real time through their respective power sensors.

[0128] In step S402, a droop control algorithm is used to calculate the reactive power regulation command. This algorithm determines the required reactive power compensation based on the voltage deviation and the current active power output value. The droop coefficient is adjusted based on the real-time ambient temperature and humidity. The droop coefficient can be adjusted using a fuzzy algorithm, allowing the droop control algorithm to adapt to environmental changes and improve control flexibility.

[0129] In step S403, the operating modes of the PV inverter and energy storage converter are adjusted based on reactive power regulation instructions. The PV inverter is set to prioritize reactive power regulation. When the PV inverter's reactive power output reaches its rated upper limit, the energy storage converter intervenes and continues to regulate reactive power, achieving coordinated control of the two, ensuring the continuity and effectiveness of reactive power compensation. This hierarchical control strategy fully utilizes the PV inverter's reactive power regulation capabilities while avoiding frequent operation of the energy storage converter, thereby extending the service life of the energy storage device.

[0130] In step S404, the operating status of the PV inverter and energy storage converter is monitored via a status monitoring module. Monitoring parameters include, but are not limited to, device temperature, current, and voltage. Overload or fault detection is accomplished by comparing the monitored parameters with preset thresholds. Once an overload or fault is detected, the backup inverter or energy storage converter is immediately activated, and its operating parameters are adjusted to take over the operation of the faulty device, ensuring voltage stability at the grid connection point. The timely intervention of backup equipment and the adjustment of operating parameters can effectively improve system reliability and fault tolerance, preventing voltage instability caused by device failure.

[0131] Specifically, in a subtropical coastal industrial park, the intelligent control system for photovoltaic energy storage power consumption maintains voltage stability at the grid connection point. First, the system uses a voltage sensor to collect real-time voltage data at the grid connection point, compares the data with a preset voltage value, and calculates the voltage deviation. Simultaneously, the system obtains the current active power output values ​​of the photovoltaic inverter and energy storage converter. Then, the system uses a droop control algorithm to calculate a reactive power regulation command based on the voltage deviation and the current active power output value. Notably, the droop coefficient in the droop control algorithm is dynamically adjusted, taking into account the real-time ambient temperature and humidity. Higher ambient temperature and humidity increase the droop coefficient adjustment, and vice versa. This allows the control system to adaptively adjust its control strategy based on environmental conditions, more effectively addressing voltage fluctuations in high-temperature and high-humidity environments. Next, the system adjusts the operating modes of the photovoltaic inverter and energy storage converter based on the calculated reactive power regulation command. Under normal circumstances, the photovoltaic inverter takes priority in reactive power regulation. When the reactive power output of the photovoltaic inverter reaches its upper limit, the energy storage converter begins to work in conjunction to jointly regulate reactive power, ensuring that the system has sufficient reactive power compensation capabilities. Finally, the system monitors the operating status of the photovoltaic inverter and energy storage converter in real time. If any device is detected to be overloaded or faulty, the system immediately activates the backup inverter or energy storage converter, adjusts the relevant operating parameters, and seamlessly switches to the backup device to ensure the continuous stability of the voltage at the grid connection point. Through the above steps, this technical solution can effectively maintain the voltage stability of the grid connection point in the high temperature and high humidity environment of subtropical coastal industrial parks.

[0132] In some preferred embodiments, step S402 includes:

[0133] According to the real-time ambient temperature and real-time humidity, the temperature correction coefficient and humidity correction coefficient are calculated by fuzzy algorithm;

[0134] Calculate the comprehensive environmental correction factor based on the temperature correction factor and the humidity correction factor;

[0135] Adjust the droop coefficient of the droop control algorithm according to the comprehensive environmental correction coefficient;

[0136] According to the voltage deviation and the current active power output value, the adjusted droop coefficient is used to calculate the reactive power regulation instructions of the photovoltaic inverter and the energy storage converter based on the droop control algorithm.

[0137] A fuzzy algorithm is used to calculate the temperature and humidity correction coefficients, addressing the nonlinear relationship between ambient temperature and humidity and the droop coefficient, enabling more refined adjustment of the droop coefficient. The temperature and humidity correction coefficients are then combined to produce a comprehensive environmental correction coefficient, fully accounting for the combined effects of temperature and humidity on the droop coefficient. This comprehensive environmental correction coefficient is then used to adjust the droop coefficient of the droop control algorithm, resulting in an adjusted droop coefficient tailored to the current environmental conditions. Finally, based on the adjusted droop coefficients, the droop control algorithm is used to calculate the reactive power regulation commands for the photovoltaic inverter and energy storage converter, making the calculation of the reactive power regulation commands more accurate and thus improving voltage stability control.

[0138] Specifically, to achieve adaptive adjustment of the droop coefficient, the real-time ambient temperature and humidity are first used as inputs to a fuzzy algorithm. After fuzzification, fuzzy inference, and defuzzification, the temperature correction coefficient and humidity correction coefficient are calculated. The fuzzy algorithm uses a pre-set fuzzy rule base that defines the value ranges of the temperature correction coefficient and humidity correction coefficient under different temperature and humidity conditions. For example, when the ambient temperature and humidity are high, the fuzzy rule base may set higher temperature correction coefficients and humidity correction coefficients, and vice versa. The temperature correction coefficient and humidity correction coefficient are combined using a weighted average to obtain a comprehensive environmental correction coefficient. In a preferred embodiment, the comprehensive environmental correction coefficient can be calculated using the following formula: Comprehensive environmental correction coefficient = w1 * temperature correction coefficient + w2 * humidity correction coefficient, where w1 and w2 are the weights of the temperature correction coefficient and humidity correction coefficient, respectively, and w1 + w2 = 1. The droop coefficient in the droop control algorithm is then adjusted based on the comprehensive environmental correction coefficient. For example, the adjusted droop coefficient can be calculated using the following formula: Adjusted droop coefficient = Original droop coefficient * Comprehensive environmental correction coefficient. Finally, the voltage deviation and the current active power output value are input into the droop control algorithm using the adjusted droop coefficient to calculate the reactive power regulation instructions for the PV inverter and energy storage converter.

[0139] In some specific embodiments, the fuzzy algorithm can employ the Mamdani fuzzy algorithm, whose fuzzy rule base can be trained and optimized offline based on historical environmental data and photovoltaic energy storage system operating data from subtropical coastal industrial parks. For example, system total reactive power demand and voltage fluctuation data under different temperature and humidity conditions can be collected. Through data analysis and expert experience, a fuzzy rule base can be established to determine the fuzzy relationships between temperature, humidity, the temperature correction coefficient, and the humidity correction coefficient. The input and output membership functions of the fuzzy controller can use triangular or Gaussian membership functions to ensure smooth and stable control. The calculation weights w1 and w2 of the comprehensive environmental correction coefficient can be adjusted based on actual operating conditions. For example, in areas where humidity has a significant impact on the system, the weight w2 of the humidity correction coefficient can be appropriately increased. The droop coefficient adjustment method can also be selected based on system characteristics, for example, different adjustment methods such as addition or multiplication can be used. Through the above-described specific embodiments, precise adaptive adjustment of the droop coefficient can be achieved, improving the accuracy and robustness of voltage control of the photovoltaic energy storage system in subtropical coastal industrial parks.

[0140] In some embodiments, step S5 includes:

[0141] S501. Real-time monitoring of the inverter temperature, energy storage battery temperature, and converter temperature of the photovoltaic energy storage system to build a device temperature data set;

[0142] S502. Based on the real-time ambient temperature, real-time humidity and device temperature data set, a fuzzy control algorithm is used to calculate the cooling demand level;

[0143] S503. According to the cooling demand level and the current operating state of the cooling system, a hierarchical control strategy is used to adjust the operating parameters of the cooling system.

[0144] In step S501, the PV energy storage system's inverter, battery, and converter temperatures are monitored in real time to construct a device temperature dataset. This is done to fully understand the operating thermal status of key PV energy storage system equipment and provide accurate baseline data for subsequent cooling control. Temperature monitoring can be achieved using temperature sensors, such as thermocouples, thermistors, or integrated temperature sensors, which can collect real-time surface temperature data from the inverter, battery, and converter. The device temperature dataset can be constructed using a data acquisition system to integrate, store, and manage the collected temperature data, providing data support for the subsequent fuzzy control algorithm.

[0145] In step S502, the calculation of the cooling demand level using a fuzzy control algorithm based on the ambient temperature data, humidity data, and device temperature data set is key to implementing the cooling system's optimized operation strategy. The application of the fuzzy control algorithm enables the system to handle the complex nonlinear relationship between ambient temperature, humidity, and device temperature, more accurately assessing actual cooling demand. The specific implementation process of the fuzzy control algorithm may include: first, determining the fuzzy input variables for ambient temperature, humidity, and device temperature, and setting the fuzzy subsets and membership functions for each variable; then, establishing fuzzy control rules. For example, when the ambient temperature, humidity, and device temperature are high, the cooling demand level is high; when the ambient temperature, humidity, and device temperature are moderate, the cooling demand level is low (the temperature and humidity can be adjusted based on actual needs); then, fuzzification processing is performed based on the real-time collected ambient temperature, humidity, and device temperature data to obtain the fuzzy membership of each input variable; finally, fuzzy inference is performed based on the fuzzy control rules, and a clear cooling demand level output value is calculated through defuzzification. The cooling demand level can be divided into multiple levels, such as low, medium, and high, or expressed in numerical values, where a higher level indicates a higher cooling demand.

[0146] In step S503, a hierarchical control strategy is employed to adjust the cooling system operating parameters based on the cooling demand level and the current operating status of the cooling system. This is intended to minimize the cooling system's energy consumption while meeting the equipment's cooling needs, thereby improving the system's overall operating efficiency and economy. The hierarchical control strategy can be implemented by presetting different cooling system operating parameters based on the cooling demand level. For example, when the cooling demand level is low, the cooling fan speed can be reduced, the number of cooling cycles can be reduced, or the cooling system can be operated intermittently. When the cooling demand level is medium, the cooling fan speed can be maintained at a medium speed, and the cooling cycle can be operated normally. When the cooling demand level is high, the cooling fan speed can be increased, the number of cooling cycles can be increased, or auxiliary cooling devices can be activated. The current operating status of the cooling system can be monitored in real time using sensors, such as fan speed sensors and coolant flow sensors, to enable precise adjustments based on the actual operating status. The specific control parameters of the hierarchical control strategy can be calibrated and optimized based on the actual application scenario and equipment characteristics.

[0147] Specifically, the cooling system optimization operation strategy proposed in this application first monitors the temperature of the key equipment of the photovoltaic energy storage system in real time through step S501 to fully grasp the thermal status of the equipment operation. Then, in step S502, the ambient temperature and humidity and the equipment temperature are comprehensively considered, and the cooling demand level is accurately calculated using the fuzzy control algorithm to achieve a refined assessment of the cooling demand. Finally, step S503 uses a hierarchical control strategy to finely adjust the cooling system operating parameters based on the cooling demand level and the current operating status of the cooling system to achieve intelligent control of the cooling system. Through the above steps, the present application can dynamically adjust the operating strategy of the cooling system according to the actual operating status and environmental conditions of the photovoltaic energy storage system, while ensuring the stable operating temperature of the equipment, reducing the energy consumption of the cooling system. Compared with the traditional fixed operating mode cooling system, the optimization strategy proposed in this application can more effectively reduce the operating temperature of the equipment and improve the overall operating efficiency of the system.

[0148] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0149] In the PV energy storage system, PT100 platinum resistance temperature sensors are installed on the inverter, energy storage battery, and converter to collect real-time temperature data from each device, which is used to construct a device temperature dataset. Temperature and humidity sensors are also installed in the PV energy storage system deployment area to collect real-time ambient temperature and humidity data. The controller uses an STM32 microcontroller with a pre-programmed fuzzy control algorithm. The algorithm takes into account three input variables: real-time ambient temperature, real-time humidity, and device temperature, and the cooling demand level as the output variable. The fuzzy control rules are set as follows: when the real-time ambient temperature is above 35°C, the real-time humidity is above 80%, and the device temperature is above 50°C, the cooling demand level is high; when the real-time ambient temperature is between 25°C and 35°C, the real-time humidity is between 50% and 80%, and the device temperature is between 40°C and 50°C, the cooling demand level is medium; and when the real-time ambient temperature is below 25°C, the real-time humidity is below 50%, or the device temperature is below 40°C, the cooling demand level is low. The cooling system uses air cooling, and the fan drive circuit uses PWM speed control. The controller outputs a PWM control signal based on the cooling demand level to adjust the fan speed. When the cooling demand level is low, the PWM duty cycle is 30%, and the fan runs at low speed; when the cooling demand level is medium, the PWM duty cycle is 60%, and the fan runs at medium speed; when the cooling demand level is high, the PWM duty cycle is 90%, and the fan runs at high speed.

[0150] refer to Figure 2 The present application also provides an industrial park photovoltaic energy storage power consumption intelligent control device for controlling power consumption of a photovoltaic energy storage system in a subtropical coastal industrial park. The device comprises:

[0151] Environmental parameter monitoring module 1 is used to obtain the real-time ambient temperature and humidity of the PV energy storage system deployment area (refer to step S1 above for the specific process);

[0152] Reactive power demand prediction module 2 is used to predict the total reactive power demand of power electronic equipment in the industrial park and photovoltaic energy storage system using a combined prediction model based on historical equipment operation data, real-time ambient temperature and real-time humidity (for details, refer to step S2 above);

[0153] Dynamic reactive power compensation control module 3, used to dynamically adjust the output of the reactive power compensation device according to the prediction results of the combined prediction model to achieve reactive power compensation (for the specific process, refer to step S3 above);

[0154] Voltage stabilization control module 4, used to adjust the operating mode of the photovoltaic inverter and energy storage converter, coordinate the output of active power and reactive power, and maintain voltage stability at the grid connection point (for the specific process, refer to step S4 above);

[0155] The equipment cooling optimization module 5 is used to optimize the operation strategy of the photovoltaic energy storage system cooling system according to the real-time ambient temperature and real-time humidity, and reduce the equipment operating temperature (for the specific process, refer to step S5 above).

[0156] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interface, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0157] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the purpose of the solution of this embodiment.

[0158] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0159] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0160] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. An intelligent control method for photovoltaic energy storage power consumption in industrial parks, used for power consumption control of photovoltaic energy storage systems in subtropical coastal industrial parks, characterized by: The method includes: S1. Obtain the real-time ambient temperature and humidity in the area where the PV energy storage system is deployed; S2. Use a combined prediction model to predict the total reactive power demand of power electronic equipment in the industrial park and the photovoltaic energy storage system based on historical equipment operating data, real-time ambient temperature, and real-time humidity. S3. According to the prediction results of the combined prediction model, dynamically adjust the output of the reactive compensation device to achieve reactive power compensation; S4. Adjust the operating modes of the photovoltaic inverter and energy storage converter to collaboratively control the output of active and reactive power to maintain voltage stability at the grid connection point; S5. Optimize the operation strategy of the photovoltaic energy storage system cooling system based on the real-time ambient temperature and humidity to reduce the equipment operating temperature; The combined prediction model includes a linear regression model and a nonlinear neural network model; the linear regression model input includes active power, ambient temperature and humidity in the historical operation data of the equipment, and the output is a first reactive power prediction value; the nonlinear neural network model input includes active power, ambient temperature and humidity in the historical operation data, and the output is a second reactive power prediction value; Step S2 includes: S201. Based on the active power, ambient temperature and humidity in the historical operation data of the device, a linear regression model is used to obtain a first reactive power prediction value, and a nonlinear neural network model is used to obtain a second reactive power prediction value; S202. Calculate a first weight of the linear regression model and a second weight of the nonlinear neural network model using a dynamic weight allocation method based on historical prediction errors of the linear regression model and the nonlinear neural network model, where the sum of the first weight and the second weight is 1; S203. An initial reactive power prediction value is obtained by weighted average calculation based on the first reactive power prediction value, the second reactive power prediction value, the first weight, and the second weight; S204. Based on the real-time ambient temperature and real-time humidity, a Sigmoid function is used to perform nonlinear correction on the initial reactive power prediction value to obtain a final reactive power prediction value.

2. The method for intelligent control of photovoltaic energy storage power consumption in an industrial park according to claim 1, characterized in that: Step S202 includes: A1. Perform sliding window processing on historical forecast errors to construct an error sequence. Then, use a wavelet decomposition algorithm to decompose the error sequence to obtain multiple subsequences of different frequencies. A2. For each subsequence, calculate the autocorrelation coefficient. Based on the autocorrelation coefficient, use a gradient descent algorithm with an adaptive step size to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. The sum of the first and second weights is 1.

3. The intelligent control method for photovoltaic energy storage power consumption in an industrial park according to claim 2, characterized in that: Step A2 includes: B1. For each subsequence, calculate the autocorrelation coefficient. If the autocorrelation coefficient is greater than a preset threshold, the corresponding subsequence is determined to be a stationary sequence and step B2 is executed. Otherwise, the corresponding subsequence is determined to be a non-stationary sequence and step B3 is executed. B2. For the stationary sequence, use a fixed-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. The fixed step size is a preset step size, which are recorded as the first weight preferred value and the second weight preferred value, respectively. B3. For non-stationary sequences, use a variable-step gradient descent algorithm to calculate the first weight of the linear regression model and the second weight of the nonlinear neural network model. These are denoted as the first preferred weight value and the second preferred weight value, respectively. The variable-step calculation formula is: st_i = st_max * exp(-|r_i|), where st_i is the step size of the i-th subsequence, st_max is the preset maximum step size, and r_i is the autocorrelation coefficient of the i-th subsequence. B4. Perform a weighted average of the first weight preference value and the second weight preference value calculated for all subsequences to obtain the final first weight of the linear regression model and the final second weight of the nonlinear neural network model. The sum of the final first weight and the final second weight is 1.

4. The method for intelligent control of photovoltaic energy storage power consumption in an industrial park according to claim 1, characterized in that: Step S3 includes: S301 real-time collection of power consumption data of various types of loads in the industrial park, identifying the first type of load that responds quickly to reactive power demand and the second type of load that responds slowly; S302. Based on the historical power consumption data ratio of the first and second types of loads, the total reactive power demand is decomposed into the reactive power demand forecast value of the first type of load and the reactive power demand forecast value of the second type of load; S303. For the reactive power demand forecast value of the first type of load, a feedforward control method is used to directly adjust the output of the reactive compensation device according to the forecast value to achieve rapid compensation; S304. Based on the reactive power demand forecast value of the second type of load, a feedback control method is adopted to monitor the grid voltage in real time. According to the grid voltage deviation and the reactive power demand forecast value of the second type of load, the output of the reactive compensation device is adjusted through the PID controller to achieve accurate compensation.

5. The method for intelligent control of photovoltaic energy storage electricity consumption in an industrial park according to claim 4, characterized in that: Step S301 includes: Real-time collection of electricity consumption data of various types of loads in the industrial park to build an electricity consumption data set; electricity consumption data includes active power, reactive power, voltage, and current; Preprocess the electricity consumption dataset to obtain a preprocessed electricity consumption dataset; the preprocessing includes missing value filling, outlier detection and elimination, and sliding average filtering; Extract the time domain features and frequency domain features of the preprocessed electricity consumption dataset. The time domain features include mean, variance, peak value and kurtosis, and the frequency domain features include total harmonic distortion rate to obtain the load feature vector. The K-means clustering algorithm is used to divide the load into the first type of load and the second type of load according to the load feature vector.

6. The method for intelligent control of photovoltaic energy storage power consumption in an industrial park according to claim 1, characterized in that: Step S4 includes: S401 real-time monitoring of the grid connection point voltage, calculate the grid connection point voltage and the preset voltage deviation, and obtain the current active power output value of the photovoltaic inverter and energy storage converter; S402. Based on the voltage deviation and the current active power output value, a droop control algorithm is used to calculate the reactive power regulation instruction of the photovoltaic inverter and the energy storage converter, wherein the droop coefficient in the droop control algorithm is adjusted according to the real-time ambient temperature and real-time humidity; S403. According to the reactive power regulation instruction, adjust the working mode of the photovoltaic inverter and the energy storage converter, and coordinate the output of active power and reactive power. Among them, the photovoltaic inverter prioritizes the adjustment of reactive power. When the reactive power of the photovoltaic inverter reaches the upper limit, the energy storage converter continues to adjust the reactive power; S404. Monitor the operating status of the photovoltaic inverter and energy storage converter in real time. If an equipment overload or failure is detected, start the backup inverter or energy storage converter and adjust the operating parameters to ensure stable voltage at the grid connection point.

7. The method for intelligent control of photovoltaic energy storage electricity consumption in an industrial park according to claim 6, characterized in that: Step S402 includes: According to the real-time ambient temperature and real-time humidity, the temperature correction coefficient and humidity correction coefficient are calculated by fuzzy algorithm; Calculate the comprehensive environmental correction factor based on the temperature correction factor and the humidity correction factor; Adjust the droop coefficient of the droop control algorithm according to the comprehensive environmental correction coefficient; According to the voltage deviation and the current active power output value, the adjusted droop coefficient is used to calculate the reactive power regulation instructions of the photovoltaic inverter and the energy storage converter based on the droop control algorithm.

8. The method for intelligent control of photovoltaic energy storage electricity consumption in an industrial park according to claim 1, characterized in that: Step S5 includes: S501. Real-time monitoring of the inverter temperature, energy storage battery temperature, and converter temperature of the photovoltaic energy storage system to build a device temperature data set; S502. Based on the real-time ambient temperature, real-time humidity and device temperature data set, a fuzzy control algorithm is used to calculate the cooling demand level; S503. According to the cooling demand level and the current operating state of the cooling system, a hierarchical control strategy is used to adjust the operating parameters of the cooling system.

9. An intelligent control device for photovoltaic energy storage electricity consumption in industrial parks, used for electricity consumption control of photovoltaic energy storage systems in subtropical coastal industrial parks, characterized by: The device includes: Environmental parameter monitoring module, used to obtain the real-time ambient temperature and humidity in the area where the photovoltaic energy storage system is deployed; The reactive power demand forecasting module is used to predict the total reactive power demand of power electronic equipment in industrial parks and photovoltaic energy storage systems using a combined forecasting model based on historical equipment operating data, real-time ambient temperature, and real-time humidity; Dynamic reactive power compensation control module, used to dynamically adjust the output of reactive power compensation device according to the prediction results of combined prediction model to achieve reactive power compensation; The voltage stabilization control module is used to adjust the operating mode of the photovoltaic inverter and energy storage converter, coordinate the output of active power and reactive power, and maintain voltage stability at the grid connection point; The equipment cooling optimization module is used to optimize the operation strategy of the photovoltaic energy storage system cooling system according to the real-time ambient temperature and real-time humidity to reduce the equipment operating temperature; The combined prediction model includes a linear regression model and a nonlinear neural network model; the linear regression model input includes active power, ambient temperature and humidity in the historical operation data of the equipment, and the output is a first reactive power prediction value; the nonlinear neural network model input includes active power, ambient temperature and humidity in the historical operation data, and the output is a second reactive power prediction value; The reactive power demand forecasting module is used to predict the total reactive power demand of power electronic equipment in industrial parks and photovoltaic energy storage systems using a combined forecasting model based on historical equipment operation data, real-time ambient temperature, and real-time humidity: S201. Based on the active power, ambient temperature and humidity in the historical operation data of the device, a linear regression model is used to obtain a first reactive power prediction value, and a nonlinear neural network model is used to obtain a second reactive power prediction value; S202. Calculate a first weight of the linear regression model and a second weight of the nonlinear neural network model using a dynamic weight allocation method based on historical prediction errors of the linear regression model and the nonlinear neural network model, where the sum of the first weight and the second weight is 1; S203. An initial reactive power prediction value is obtained by weighted average calculation based on the first reactive power prediction value, the second reactive power prediction value, the first weight, and the second weight; S204. Based on the real-time ambient temperature and real-time humidity, a Sigmoid function is used to perform nonlinear correction on the initial reactive power prediction value to obtain a final reactive power prediction value.

Images