Control method of photovoltaic direct-drive air source heat pump unit

Abstract

The invention relates to the technical field of heat pump unit control, in particular to a control method of a photovoltaic direct-driven air source heat pump unit, which comprises the following steps of: acquiring historical environment data of the heat pump unit according to a sampling frequency, performing trend identification based on the historical environment data, acquiring a predicted value of the environment data at the current moment, and calculating the predicted value of the environment data; obtaining a power generation power trend value at the current moment; according to the actual value and the trend value of the power generation power at the current moment, obtaining the initial trend identification accuracy and the confidence coefficient of the next moment, and obtaining the target trend identification accuracy of the next moment according to the accuracy and the confidence coefficient; and sequentially adjusting the initial trend identification windows, and finally controlling the unit by using an MPC algorithm according to the adjusted trend identification windows. According to the method, the trend identification accuracy and confidence are calculated through the difference between the trend identification result and the actual data, and the trend identification window is adjusted based on the confidence and the accuracy, so that the accurate trend identification result is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of heat pump unit control, and in particular to a control method for a photovoltaic direct-driven air source heat pump unit. Background Art

[0002] As the global energy structure accelerates its transformation toward a low-carbon one, severe challenges are posed to traditional energy utilization models. As an innovative system for the coordinated utilization of "light-heat-electricity", the photovoltaic direct-driven air source heat pump unit has become the core technology path for achieving low-carbon transformation in building heating, industrial heating and other fields by virtue of its unique advantages. The system uses photovoltaic components to directly convert solar energy into electrical energy to drive the heat pump, effectively avoiding the multi-stage conversion losses of traditional power grid power supply. In theory, it can achieve near-zero-carbon heating, combining environmental value with economic potential. However, with the complexity of the system operation scenario, how to achieve precise control to improve operating efficiency has become a key issue restricting its large-scale promotion.

[0003] At present, using data trend recognition technology to optimize model prediction is an important direction to improve the control accuracy of photovoltaic heat pump units. Existing technologies mostly use MPC models to predict and control unit data. In this process, trend recognition methods with fixed window sizes expose significant defects. On the one hand, fixed windows are difficult to adapt to the complex and changeable trend characteristics in data sequences, and are prone to miss subtle changes in unclear trends, resulting in the loss of key information and the inability to fully capture data dynamics; on the other hand, for areas with large fluctuations, fixed windows will over-enhance local trend characteristics, resulting in deviations and misinterpretations of trend judgments. This will lead to a decrease in the accuracy of the MPC prediction model, and make it difficult for the system to make timely and effective control responses when facing complex working conditions such as sudden changes in light intensity and drastic fluctuations in ambient temperature, thereby affecting the economy, reliability and low-carbon benefits of unit operation. Summary of the invention

[0004] To solve the technical problem that the trend recognition method using a fixed window size affects the accuracy of the MPC prediction model, the present invention provides a control method for a photovoltaic direct-drive air source heat pump unit; the method includes the steps of: collecting the actual historical power generation power of the heat pump unit according to a preset sampling frequency, and historical environmental data including multiple environmental parameters such as ambient temperature and wind speed; setting an initial trend recognition window, performing trend recognition based on the historical environmental data to obtain a predicted value of the environmental data at the current moment; calculating a power generation power trend value at the current moment through a photovoltaic power generation power model based on the predicted value of the environmental data at the current moment; obtaining the initial trend recognition accuracy and its confidence level of the power generation power at the next moment according to the actual power generation power value and the power generation power trend value at the current moment, and obtaining the target trend recognition accuracy of the power generation power at the next moment according to the initial trend recognition accuracy and its confidence level; adjusting the initial trend recognition window according to the target trend recognition accuracy to obtain a target trend recognition window; and controlling the heat pump unit by using the MPC algorithm according to the target trend recognition window.

[0005] The present invention combines dynamic adjustment of the trend recognition window with the MPC algorithm to achieve precise control of the photovoltaic direct-drive heat pump unit. Specifically, after obtaining the power generation power trend value based on the sampled data, the initial trend recognition accuracy and confidence level are calculated through the difference between the actual value and the trend value, and then the target trend recognition accuracy is obtained, so as to dynamically optimize the size of the trend recognition window. This mechanism can adapt to fluctuations in environmental data, expand the window to capture stable trends when the data changes violently, and shrink the window to improve the calculation efficiency when the data is stable, avoiding the limitations of a fixed window. Combining with the rolling optimization characteristics of the MPC algorithm, it can respond to changes in photovoltaic power and environmental parameters in real time, precisely adjust the heat pump operation strategy, improve the "light-thermal-electricity" synergy efficiency, reduce equipment losses and energy consumption caused by prediction deviations, enhance the adaptability of the system to complex working conditions, and achieve efficient and low-carbon operation.

[0006] Preferably, the step of obtaining the initial trend recognition accuracy and its confidence level of the power generation power at the next moment according to the actual power generation power value and the power generation power trend value at the current moment includes: obtaining the actual power generation power values at the current moment and the previous several moments, denoted as power generation power neighborhood data; obtaining the initial trend recognition accuracy of the power generation power at the next moment according to the relative deviation between the actual power generation power value and the power generation power trend value of the power generation power neighborhood data, and the fluctuation degree of the actual power generation power value in the power generation power neighborhood data; and obtaining the confidence level of the initial trend recognition accuracy at the next moment according to the actual power generation power value at the next moment and the power generation power neighborhood data.

[0007] Based on the gap between the actual power generation value and the trend value, the present invention quantifies the initial accuracy, combines the fluctuation degree of neighborhood data to correct error interference, and can effectively filter the influence of short-term outliers on trend recognition. At the same time, by obtaining the confidence through the correlation of neighborhood actual values, the stability of the data sequence can be dynamically evaluated. This mechanism enables the system to automatically reduce the weight of unreliable trend recognition results and avoid misjudgment when the environment changes suddenly, and strengthen the effectiveness of trend prediction during the data stable period, helping the MPC algorithm to more accurately adjust the operating parameters of the heat pump.

[0008] Preferably, the obtaining of the initial trend recognition accuracy at the next moment includes: denoting the next moment as the -th moment, and the initial trend recognition accuracy at the next moment satisfies the relational expression: ; where is the initial trend recognition accuracy of the power generation at the -th moment, is the natural exponential function, is the difference between the index of the power generation trend value at the -th moment and the index of the -th power generation neighborhood data, is the relative deviation between the actual power generation value and the power generation trend value of the -th power generation neighborhood data of the power generation at the -th moment, is the influence factor of the fluctuation degree of the actual power generation value in the power generation neighborhood data, is the length of the preset trend recognition window, and the -th power generation neighborhood data of the power generation at the -th moment is the data in its preset trend recognition window.

[0009] The present invention correlates the difference between neighborhood data indexes and the actual trend difference through the natural exponential function, excavates the local fluctuation characteristics of the power sequence, and has strong anti-noise performance; incorporates the influence factor of the fluctuation of the actual power generation value to adapt to environmental changes and enhance robustness; combines the preset window length to integrate local and global trends, provides a reliable basis for the MPC algorithm to regulate the unit, and improves the energy efficiency and stability of the system.

[0010] Preferably, the relative deviation between the actual power generation value and the power generation trend value of the -th power generation neighborhood data of the power generation at the -th moment satisfies the relational expression: ; where is the power generation trend value of the -th power generation neighborhood data of the power generation trend value at the -th moment, is the The actual power generation value of the power generation power neighborhood data at a moment, is the absolute value.

[0011] Preferably, the influence factor of the fluctuation degree of the actual power generation value in the power generation power neighborhood data satisfies the relational expression: ; where is the standard deviation of the power generation power neighborhood data of the th actual power generation value among the th environmental parameters, is the average value of the power generation power neighborhood data of the th actual power generation value among the th environmental parameters, is a small value used to prevent the denominator from being zero, is the total number of items of environmental parameters in each environmental data.

[0012] Preferably, the confidence level of the initial trend recognition accuracy at the next moment satisfies the relational expression: ; where is the confidence level of the initial trend recognition accuracy at the th moment, is the length of the preset trend recognition window, is the th actual power generation value of the th power generation power neighborhood data of the power generation power trend value at the th moment, is the actual power generation value at the th moment, is the range of the actual power generation values at multiple moments,

[0013] The present invention constructs a model based on the deviation between the neighborhood actual power value and the actual value at the current moment, and combines the natural exponential function. By associating neighborhood data through the preset window length , and normalizing the deviation with the range of the actual power generation value, the credibility of the initial trend recognition accuracy can be accurately quantified. It not only adapts to the data fluctuation characteristics, but also strengthens the influence of the deviation on the confidence level through exponential operation, provides a reliable confidence evaluation for subsequent trend recognition window adjustment and unit control optimization, and improves the adaptation and regulation accuracy of the system to complex working conditions.

[0014] Preferably, the target trend recognition accuracy is the product of the initial trend recognition accuracy of the power generation power at the next moment and its confidence level.

[0015] Preferably, the initial trend recognition window is set, and trend recognition is performed based on the historical environment data to obtain the predicted value of the environment data at the current moment, including: obtaining multiple environmental parameters in the historical environment data to form the time series data of each environmental parameter, and taking the data points in the initial trend recognition window of each data point in the time series data corresponding to each environmental parameter as the environmental parameter neighborhood data; performing curve fitting according to the environmental parameters at each moment in the environmental parameter neighborhood data to obtain the fitting slope at the current moment, and predicting the environmental parameter at the current moment according to the fitting slope to obtain the predicted value of the environment data at the current moment.

[0016] Preferably, based on the predicted value of the environment data at the current moment, the power generation power trend value at the current moment is calculated through the photovoltaic power generation power model, including: constructing a photovoltaic power generation power prediction model according to the heat pump COP function and combining the environment data at the current moment to obtain the power generation power trend value at the current moment.

[0017] Preferably, the target trend recognition window satisfies the relational expression: ; where is the length of the target trend recognition window, is the length of the preset trend recognition window, is the target trend recognition accuracy of the power generation power at the th moment, is the maximum value function, is the decimal of, used to limit the minimum value.

[0018] Based on the preset window length, the present invention dynamically adjusts the window through the target trend recognition accuracy. It not only adapts the window size according to the accuracy, but also borrows limitations to avoid excessive contraction of the window, ensure data reliability, and make the trend recognition window adapt to the working condition changes.

[0019] Advantages of the present invention: The present invention realizes precise control of a photovoltaic direct-drive heat pump unit through a multi-dimensional strategy. By combining the difference between the actual power generation value and the power generation trend value, the initial trend recognition accuracy and confidence are calculated, and then the target accuracy is deduced, thereby dynamically optimizing the trend recognition window to obtain a more accurate trend recognition result. This window can adapt to fluctuations in environmental data, expanding when the data changes violently to capture stable trends and shrinking when stable to improve calculation efficiency, breaking through the limitations of a fixed window. By using the natural exponential function to correlate neighborhood data, local fluctuation characteristics of the power generation sequence are mined to enhance noise resistance; by incorporating the influence factor of the actual value fluctuation of the power generation, it adapts to environmental changes and improves robustness. At the same time, based on the deviation between the neighborhood and the actual power generation value at the current moment, a confidence model is constructed using the natural exponential function and range normalization to accurately quantify the credibility of the initial trend recognition accuracy, providing a reliable basis for adjusting the trend recognition window and optimizing the unit control. Description of the Drawings

[0020] Figure 1 It is a flowchart of a control method for a photovoltaic direct-drive air-source heat pump unit provided by an embodiment of the present invention. Detailed Embodiments

[0021] An embodiment of the present invention provides a control method for a photovoltaic direct-drive air-source heat pump unit, as Figure 1 shown. The method includes steps S100 - step S600: Step S100, collect the historical actual power generation value of the heat pump unit and historical environmental data including multiple environmental parameters such as environmental temperature and wind speed according to a preset sampling frequency.

[0022] It should be noted that during the operation of the photovoltaic direct-drive air-source heat pump unit, the power generation efficiency of the photovoltaic module is affected by various natural environmental data, such as solar irradiance intensity, environmental temperature, humidity, wind speed, etc. Solar irradiance intensity determines the photoelectric conversion basis of the photovoltaic module. In actual operation, its value fluctuates violently from early morning to noon and is significantly positively correlated with the power generation efficiency. Environmental temperature is negatively correlated with the power generation efficiency. As the temperature rises, the power generation efficiency of silicon-based photovoltaic modules usually shows a downward trend. The influence of humidity on the power generation efficiency is twofold. On the one hand, a high-humidity environment will cause dew condensation on the surface of the module, forming water stains to block light; on the other hand, humidity will accelerate the adsorption of dust on the surface of the module. When the dust accumulation causes the light transmittance to decrease, the power generation efficiency will be significantly inhibited. Although wind speed does not directly participate in photoelectric conversion, it can reduce the module temperature through convective heat dissipation, thereby improving the power generation efficiency. The real-time changes of these environmental data act together on the power generation efficiency of the photovoltaic module. Therefore, collecting and analyzing them is the key to optimizing the operation control of the photovoltaic direct-drive air-source heat pump unit.

[0023] Specifically, set the sampling frequency to once every 10 minutes, set the sampling points, and install devices such as temperature sensors, humidity sensors, irradiance sensors, and wind speed sensors at the sampling points to collect environmental data such as ambient temperature, humidity, solar irradiance intensity, and wind speed respectively. Denote the current sampling moment as the current moment, and several moments before the current moment as historical moments. After the above-mentioned collection method, data on ambient temperature, humidity, solar irradiance intensity, and wind speed at several historical moments will be obtained, that is, historical environmental data.

[0024] In addition, it is also necessary to set sensors built into the photovoltaic busbar box or inverter to collect the actual power generation data of the unit. The sampling frequency of this data is consistent with that of the environmental data, and this data is denoted as the actual value of the power generation. Similarly, the actual values of the power generation at several historical moments can be obtained.

[0025] So far, several historical actual values of the power generation are obtained, as well as historical environmental data including multiple environmental parameters such as ambient temperature and wind speed. In addition, the actual value of the power generation and environmental data at the current moment can also be obtained.

[0026] Step S200: Set an initial trend recognition window, perform trend recognition based on the historical environmental data, and obtain the predicted value of the environmental data at the current moment.

[0027] It should be noted that in order to obtain the trend value of the power generation, it is necessary to first obtain the predicted value of the environmental data, and then calculate the trend value of the power generation using the power generation prediction model based on the predicted values of each environmental data.

[0028] For the predicted value of each environmental parameter, curve fitting or exponential smoothing can be used. Taking curve fitting as an example, curve fitting can effectively explore the potential non-linear change rules in the time series. Compared with simple methods such as mean and difference, it can more accurately depict the complex fluctuation trend of environmental variables. Based on the fitting slope for trend prediction, it can dynamically reflect the change rate and direction of the data. When the environmental conditions change suddenly, such as a sudden increase in wind speed or a sharp drop in temperature, the real-time update of the slope can quickly capture the trend change, enabling the prediction model to adapt to the new working conditions in a timely manner, and improving the response speed and prediction accuracy to the dynamic changes of the environment. Therefore, the present invention performs trend recognition through curve fitting. According to the trend prediction, more accurate predicted values of each environmental parameter can be obtained, and combining the trend prediction for the prediction of photovoltaic power generation can obtain more accurate and long-term photovoltaic power generation prediction results.

[0029] Specifically, first, set the length of the initial trend recognition window to Secondly, each environmental parameter in the historical environmental data obtained through the above steps constitutes a time series data sequence of the corresponding environmental parameter, and this sequence is arranged in order of time stamps. Then, each data point in the initial trend recognition window of the time series data corresponding to each environmental parameter at each historical sampling moment is obtained, and data points are recorded as its environmental data neighborhood data. Then, curve fitting is performed according to the environmental parameters at each moment in the environmental parameter neighborhood data to obtain the fitting slope at the current moment. Then, data prediction is performed according to the fitting slope to obtain the trend value of the predicted value of each environmental parameter in the environmental data at the current moment. Curve fitting and exponential smoothing are existing technologies and will not be elaborated here. The setting of the length of the initial trend recognition window can be set according to requirements. Then, each data point in the initial trend recognition window of the time series data corresponding to each environmental parameter at each historical sampling moment is obtained, and data points are recorded as its environmental data neighborhood data. Then, curve fitting is performed according to the environmental parameters at each moment in the environmental parameter neighborhood data to obtain the fitting slope at the current moment. Then, data prediction is performed according to the fitting slope to obtain the trend value of the predicted value of each environmental parameter in the environmental data at the current moment. Curve fitting and exponential smoothing are existing technologies and will not be elaborated here. The setting of the length of the initial trend recognition window can be set according to requirements.

[0030] For example, set the length of the initial trend recognition window to 3. Taking the environmental temperature as an example to illustrate the acquisition of neighborhood data, assume that the time series data of the predicted value of the environmental temperature obtained according to the above steps is Taking the third sampling moment point as an example, its environmental data neighborhood data is .

[0031] So far, the predicted value of the environmental data at the current moment has been obtained Step S300: Based on the predicted value of the environmental data at the current moment, calculate the trend value of the power generation power at the current moment through the photovoltaic power generation power model.

[0032] It should be noted that the MPC algorithm is a control strategy. By establishing a system mathematical model, it predicts the future behavior of the system based on current and historical data. At each sampling moment, the MPC calculates the control sequence for a period of time in the future according to the prediction model to minimize the objective function, and applies the first control quantity to the system. This process is repeated at the next moment to achieve rolling optimization. The COP function is a key indicator to measure the energy conversion efficiency of the heat pump system. The higher the COP value, the more heat or energy the heat pump system can output under the same power consumption, and the higher the energy efficiency. The COP function is usually related to factors such as environmental temperature and heat load demand. In the present invention, the MPC prediction algorithm is used to predict the current photovoltaic power generation efficiency data, and the difference of the subsequent data is obtained according to the predicted value.

[0033] Specifically, according to the heat pump COP function, combined with the predicted values of the historical environmental data and the current environmental data, a photovoltaic power generation power prediction model is constructed for photovoltaic power generation power prediction, and the predicted value of the photovoltaic power generation power at the current sampling moment is obtained, which is recorded as the trend value of the power generation power at the current moment.

[0034] So far, the trend value of the power generation power at the current moment has been obtained.

[0035] Step S400: Obtain the initial trend recognition accuracy and its confidence level of the power generation at the next moment according to the actual value of the power generation at the current moment and the power generation trend value, and obtain the target trend recognition accuracy of the power generation at the next moment according to the initial trend recognition accuracy and its confidence level.

[0036] It should be noted that according to the above steps, the power generation trend value is obtained through methods such as curve fitting and trend prediction. This power generation trend value reflects the changing trend of the power generation predicted based on historical data and environmental variables, providing an important reference basis for the operation control and optimization of the system. However, the accuracy of the power generation trend value is directly related to the effectiveness of subsequent control strategies. Therefore, whether it is accurate still needs to be compared and evaluated with the actual power generation.

[0037] Specifically, obtain the actual values of the power generation at the current moment and the previous moments, which are recorded as the power generation neighborhood data; the present invention evaluates the accuracy of the power generation trend value at the next moment according to the gap between the power generation trend value and the actual value of the power generation at the current moment. Since the data needs to be further adjusted later, the accuracy of the power generation trend value at the next moment obtained in this step is recorded as the initial trend recognition accuracy at the next moment. Denote the next moment as the th moment. The initial trend recognition accuracy at the next moment satisfies the expression: ; where is the initial trend recognition accuracy of the power generation at the th moment, is the natural exponential function, is the difference between the index of the power generation trend value at the th moment and the index of the th power generation neighborhood data, is the relative deviation between the actual value of the power generation of the th power generation neighborhood data and the power generation trend value at the th moment, is the influence factor of the fluctuation degree of the actual value of the power generation in the power generation neighborhood data, is the length of the preset trend recognition window. The th power generation neighborhood data of the power generation at the th moment is the data in its preset trend recognition window.

[0038] In this formula, is the th power generation of the The relative deviation between the actual power generation value and the power generation trend value of the power generation power neighborhood data. The larger this relative deviation is, the less accurate the result of trend recognition is. This difference satisfies the relational expression: ; Wherein, is the relative deviation between the actual power generation value and the power generation trend value of the th power generation power neighborhood data at the th moment, is the power generation trend value of the th power generation power neighborhood data at the th moment of the power generation trend value, is the actual power generation value of the th power generation power neighborhood data at the th moment of the power generation trend value, is the absolute value.

[0039] In the formula, is the influence factor of the fluctuation degree of the actual power generation value. The larger this value is, the greater the fluctuation degree of this part of the data, and the greater its influence on trend recognition, and the accuracy of its trend recognition will also be reduced accordingly. The influence factor of the fluctuation degree of the actual power generation value satisfies the relational expression: ; Wherein, is the influence factor of the fluctuation degree of the actual power generation value in the power generation power neighborhood data, is the th standard deviation of the power generation power neighborhood data of the rd actual power generation value in the th environmental parameter item, is the th average value of the power generation power neighborhood data of the th actual power generation value in the th environmental parameter item, is a small value used to prevent the denominator from being zero,

[0040] The above text illustrates the calculation method of the initial trend recognition accuracy for the next moment. The following text illustrates how to calculate the confidence level of the initial trend recognition accuracy for the next moment.

[0041] It should be noted that for the initial trend recognition accuracy of any item of data at the next moment, if the trend change of the data itself is not obvious, the accuracy calculation of its trend recognition also needs to be corrected.

[0042] Specifically, the confidence level of the initial trend recognition accuracy at the next moment satisfies the relational expression: ; wherein, is the confidence level of the initial trend recognition accuracy at the -th moment, is the length of the preset trend recognition window, is the actual power generation value of the -th power generation power neighborhood data of the power generation power trend value at the -th moment, is the actual power generation value at the -th moment, is the range of the actual power generation values at multiple moments, is the natural exponential function.

[0043] The part in this formula is the gap between the actual value of the power generation power data and its neighborhood data. The volatility is reflected by the gap between the neighborhood power generation power data, and the confidence level of the trend recognition accuracy is calculated. If the volatility of the power generation power data is greater, it indicates that its trend is not obvious, and then the confidence level of its trend recognition result is lower.

[0044] Thus far, the initial trend recognition accuracy and its confidence level at the next moment are obtained according to the above operations. Next, it is explained how to obtain the target trend recognition accuracy at the next moment based on these two data. The target trend recognition accuracy at the next moment satisfies the relational expression: ; wherein, is the target trend recognition accuracy of the power generation power at the -th moment, is the initial trend recognition accuracy of the power generation power at the -th moment; is the confidence level of the initial trend recognition accuracy of the power generation power at the -th moment.

[0045] In this formula, the confidence level reflects the degree of trust in the initial trend recognition accuracy. Through the product of the two, the final accuracy can be dynamically adjusted according to factors such as data stability and model credibility. For example, when the environmental data fluctuates violently or the model is in the parameter calibration stage, a lower confidence level will reduce the target accuracy, avoiding decision-making mistakes caused by over-relying on unreliable predictions.

[0046] Thus far, the target trend recognition accuracy at the next moment is obtained.

[0047] Step S500: Adjust the initial trend recognition window according to the target trend recognition accuracy to obtain the target trend recognition window.

[0048] It should be noted that for data with a greater trend recognition accuracy, the trend recognition performed is more accurate, and a smaller window can be used for trend recognition to reduce the data calculation amount. For areas with a lower trend recognition accuracy, a larger window size should be used for trend recognition to obtain a more accurate trend recognition result.

[0049] Specifically, obtain the adaptive trend recognition window size of each environmental parameter at the next moment based on the ratio of the preset trend recognition window size to the adjusted trend recognition accuracy, denoted as the target trend recognition window. The target trend recognition window satisfies the relational expression: ; where, is the length of the target trend recognition window, is the length of the preset trend recognition window, is the th target trend recognition accuracy at the moment, is the maximum value function, is a decimal number used to limit the minimum value.

[0050] In this formula, by limiting the minimum value to a decimal number between 0 and 1, such as 0.5, it is ensured that the trend recognition window will not be too large and distorted. In actual use, it can be set according to the actual implementation situation .

[0051] Thus, the target trend recognition window of each environmental parameter at the next moment is obtained.

[0052] Step S600: Control the heat pump unit using the MPC algorithm according to the target trend recognition window.

[0053] The target trend recognition window obtained according to the above steps is no longer a window with a fixed size, which can solve the technical problem that the existing trend recognition method using a fixed window size affects the accuracy of the MPC prediction model. Other control operations after obtaining the target trend recognition window are the same as those of the existing MPC algorithm and will not be elaborated here.

[0054] The above are all preferred embodiments of the present invention. The protection scope of the present invention is not limited thereby. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A control method for a photovoltaic direct-drive air source heat pump unit, characterized in that Including the steps: Collect the actual historical power generation value of the heat pump unit according to a preset sampling frequency, and historical environmental data including multiple environmental parameters such as ambient temperature and wind speed; Set an initial trend recognition window, perform trend recognition based on the historical environmental data, and obtain the predicted value of the environmental data at the current moment; Based on the predicted value of the environmental data at the current moment, calculate the power generation power trend value at the current moment through a photovoltaic power generation model; According to the actual power generation value and the power generation power trend value at the current moment, obtain the initial trend recognition accuracy and its confidence level of the power generation power at the next moment, and obtain the target trend recognition accuracy of the power generation power at the next moment according to the initial trend recognition accuracy and its confidence level; Adjust the initial trend recognition window according to the target trend recognition accuracy to obtain a target trend recognition window; According to the target trend recognition window, control the heat pump unit using the MPC algorithm.

2. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 1, wherein The obtaining of the initial trend recognition accuracy and its confidence level of the power generation power at the next moment according to the actual power generation value and the power generation power trend value at the current moment includes: Obtain the actual power generation value at the current moment and the previous several moments, which is recorded as the neighborhood data of power generation Obtain the initial trend recognition accuracy of the power generation power at the next moment according to the relative deviation between the actual power generation value and the power generation power trend value of the power generation power neighborhood data, and the fluctuation degree of the actual power generation value in the power generation power neighborhood data; Obtain the confidence level of the initial trend recognition accuracy at the next moment according to the actual power generation value at the next moment and the power generation power neighborhood data.

3. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 2, characterized in that, The obtaining of the initial trend recognition accuracy at the next moment includes: Denote the next moment as the th moment, and the initial trend recognition accuracy of the next moment satisfies the relational expression: ; Among them, is the initial trend recognition accuracy of the power generation power at the th moment, is the natural exponential function, is the difference between the index of the power generation power trend value at the th moment and the index of the th neighborhood data of the power generation power, is the relative deviation between the actual power generation power value and the power generation power trend value of the th neighborhood data of the power generation power at the th moment, is the influence factor of the fluctuation degree of the actual power generation power value in the neighborhood data of the power generation power, is the length of the preset trend recognition window, and the th neighborhood data of the power generation power at the th moment is the data in its preset trend recognition window.

4. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 3, characterized in that, The actual power generation value of the th neighborhood data of the power generation power at the th moment and the relative deviation between the power generation power trend value satisfy the relational expression: ; Among them, is the power generation power trend value of the th power generation power neighborhood data at the th moment, is the actual power generation power value of the th power generation power neighborhood data at the th moment, is the absolute value.

5. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 3, characterized in that The influence factor of the fluctuation degree of the actual power generation value in the neighborhood data of the power generation power satisfies the relational expression: ; Among them, is the standard deviation of the power generation neighborhood data of the th actual power generation value among the th environmental parameters, is the average value of the power generation neighborhood data of the th actual power generation value among the th environmental parameters, is a small value used to prevent the denominator from being zero, is the total number of environmental parameters in each environmental data.

6. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 2, characterized in that, The confidence level of the initial trend recognition accuracy at the next moment satisfies the relational expression: ； Among them, is the confidence of the initial trend recognition accuracy at the th moment, is the length of the preset trend recognition window, is the th actual power generation value of the th power generation neighborhood data of the power generation power trend value at the th moment, is the actual power generation value at the th moment, is the natural exponential function.

7. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 1, characterized in that The target trend recognition accuracy is the product of the initial trend recognition accuracy of the power generation power at the next moment and its confidence level.

8. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 1, wherein The setting of the initial trend recognition window, performing trend recognition based on the historical environmental data, and obtaining the predicted value of the environmental data at the current moment includes: Obtain multiple environmental parameters in historical environmental data to form time-series data for each environmental parameter, and record each data point in the initial trend recognition window of the time-series data corresponding to each environmental parameter as environmental parameter neighborhood data; data points are denoted as environmental parameter neighborhood data; Perform curve fitting according to the environmental parameters at each moment in the environmental parameter neighborhood data, obtain the fitting slope at the current moment, predict the environmental parameters at the current moment according to the fitting slope, and obtain the predicted value of the environmental data at the current moment.

9. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 1, characterized in that The calculating of the power generation power trend value at the current moment through a photovoltaic power generation model based on the predicted value of the environmental data at the current moment includes: According to the heat pump COP function, construct a photovoltaic power generation prediction model in combination with the environmental data at the current moment, and obtain the power generation power trend value at the current moment.

10. The control method of the photovoltaic direct-drive air source heat pump unit according to claim 1, characterized in that, The target trend recognition window satisfies the relational expression: ; Among them, is the length of the target trend recognition window, is the length of the preset trend recognition window, is the -th moment's target trend recognition accuracy of the power generation power, is the maximum value function, is 's decimal, used to limit the minimum value.

Images