Method, device and electronic equipment for predicting power regulation space of cold station

Abstract

The application relates to the technical fields of cold station regulation and control and artificial intelligence, and discloses a method and device for predicting the power regulation space of a cold station and electronic equipment, the method comprising: acquiring future weather data, chilled water temperature data and chilled water flow data of the cold station; predicting the required cold quantity data of the cold station based on the future weather data, the chilled water temperature data and the chilled water flow data; obtaining the power data of the cold station based on the required cold quantity data and system efficiency data of the cold station; and obtaining the power regulation space of the cold station based on the power data of the cold station. Through the above method, the accuracy of the prediction of the power regulation space of the cold station can be improved, and the energy utilization rate can be effectively improved under the premise of ensuring the refrigeration requirement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of cold station regulation, artificial intelligence, and the like, and in particular to a method and device for predicting the power regulation space of a cold station and an electronic device. BACKGROUND

[0002] With the increasing demand for air conditioning and refrigeration in large public buildings and industrial production, the proportion of the chilled water unit and its auxiliary equipment (cooling tower, water pump, etc.) of the air conditioning system in the total energy consumption continues to rise. The temperature and flow rate of the chilled water side and the cooling water side directly determine the operating load and efficiency of the unit.

[0003] However, in actual engineering, the regulation of the power of the cold station system is affected by various complex factors, including weather data (such as ambient temperature, humidity, air pressure, etc.), building load fluctuations (such as personnel density, equipment heat dissipation, and diurnal cycle changes), system efficiency of the cold station (such as cooling tower and water pump efficiency, chiller efficiency), and unit-specific characteristics (such as part-load characteristics, start-stop strategy). The dynamic changes of these factors make the regulation requirements of the cold station power complex and changeable, and if the adjustable power space of the cold station as a whole cannot be evaluated in real time and accurately, the energy-saving potential cannot be fully tapped. Therefore, there is an urgent need for a method for predicting the power regulation space of a cold station, which can accurately predict and evaluate the power regulation space of the chiller and supporting facilities in real time, and improve the energy utilization efficiency under the premise of ensuring the refrigeration demand. SUMMARY

[0004] The embodiments of the present application aim to at least solve one of the technical problems in the related art. To this end, the embodiments of the present application propose a method and device for predicting the power regulation space of a cold station, an electronic device, a computer product, and a medium.

[0005] The embodiments of the present application provide a method for predicting the power regulation space of a cold station, the method comprising: obtaining future weather data, chilled water temperature data, and chilled water flow rate data of the cold station; predicting demand cooling capacity data of the cold station based on the future weather data, the chilled water temperature data, and the chilled water flow rate data; obtaining cold station power data based on the demand cooling capacity data and system efficiency data of the cold station; and obtaining the power regulation space of the cold station based on the cold station power data.

[0006] In some embodiments, the demand cooling capacity data comprises a demand cooling capacity maximum value and a demand cooling capacity minimum value; the cold station power data comprises a cold station power maximum value and a cold station power minimum value; the cold station power data is obtained based on the demand cooling capacity data and system efficiency data of the cold station, comprising: the cold station power maximum value is obtained based on the demand cooling capacity maximum value and the system efficiency data; the cold station power minimum value is obtained based on the demand cooling capacity minimum value and the system efficiency data; wherein the power adjustment space of the cold station is obtained based on the cold station power data, comprising: the difference between the cold station power maximum value and the cold station power minimum value is determined as the power adjustment space.

[0007] In some embodiments, the method further comprises: obtaining system efficiency data of the cold station; obtaining the system efficiency data of the cold station, comprising: obtaining historical demand cooling capacity data and cooling water temperature data; obtaining initial system efficiency data based on the historical demand cooling capacity data by querying a table, wherein the table comprises a plurality of system efficiency values corresponding to a plurality of cooling capacity gradient values; adjusting the initial system efficiency data based on the future weather data and the cooling water temperature data to obtain the system efficiency data.

[0008] In some embodiments, the initial system efficiency data is adjusted based on the future weather data and the cooling water temperature data to obtain the system efficiency data, comprising: inputting the future weather data into a cooling water temperature prediction model to obtain the cooling water temperature data; inputting the cooling water temperature data into a system efficiency prediction model to obtain an adjustment parameter; adjusting the initial system efficiency data based on the adjustment parameter to obtain the system efficiency data.

[0009] In some embodiments, the demand cooling capacity data of the cold station is predicted based on the future weather data, the chilled water temperature data and the chilled water flow data, comprising: inputting the future weather data, the chilled water temperature data and the chilled water flow data into a cooling capacity prediction model for prediction, and outputting the demand cooling capacity data of the cold station; the cooling capacity prediction model is obtained by training in the following manner: obtaining first sample data, wherein the first sample data comprises weather sample data, chilled water supply temperature sample data, chilled water return temperature sample data and chilled water flow sample data; training the cooling capacity prediction model based on the first sample data.

[0010] In some embodiments, the cooling capacity prediction model comprises a reinforcement learning model; the cooling capacity prediction model is trained based on the first sample data, comprising: constructing a state space based on the first sample data; obtaining an action space based on the state space, wherein the action space comprises the predicted demand cooling capacity data; determining a reward parameter for the action space based on a reward function, wherein the reward parameter is associated with the chilled water return temperature; updating model parameters of the reinforcement learning model based on the reward parameter to obtain a trained reinforcement learning model.

[0011] In some embodiments, based on the reward parameter, the model parameters of the reinforcement learning model are updated to obtain a trained reinforcement learning model, including: caching the current state, the demand cold quantity data corresponding to the current action, the current reward parameter and the next state; sampling from the cached data, and updating the model parameters of the reinforcement learning model according to the time difference error mode.

[0012] In some embodiments, the chilled water return water temperature associated with the reward parameter is obtained by: inputting the weather sample data of the cold station, the chilled water supply water temperature sample data and the chilled water flow sample data into an environment simulator for prediction to obtain the chilled water return water temperature.

[0013] In some embodiments, the cooling water temperature prediction model includes a multi-layer fully connected neural network, and the cooling water temperature model is trained by: obtaining second sample data and cooling water temperature labels corresponding to the second sample data, wherein the second sample data includes historical outdoor temperature, historical outdoor humidity, historical wind speed and historical solar radiation; processing the cooling water temperature sample data by using the multi-layer fully connected neural network to obtain cooling water temperature data; determining the error between the cooling water temperature prediction data and the cooling water temperature label by using the root mean square error mode, and adjusting the model parameters of the cooling water temperature prediction model based on the error to obtain a trained cooling water temperature prediction model.

[0014] In some embodiments, the system efficiency prediction model includes a multi-layer fully connected neural network, and the system efficiency prediction model is trained by: obtaining third sample data and system efficiency labels corresponding to the third sample data, wherein the third sample data includes historical cooling water temperature data; processing the third sample data by using the multi-layer fully connected neural network to obtain system efficiency prediction data; determining the error between the system efficiency prediction data and the system efficiency label by using the root mean square error mode, and adjusting the model parameters of the system efficiency model based on the error to obtain a trained system efficiency prediction model.

[0015] Embodiments of the present application provide a device for predicting the power adjustment space of a cold station, the device comprising: an obtaining module configured to obtain future weather data, chilled water temperature data and chilled water flow data of the cold station; a prediction module configured to predict demand cold quantity data of the cold station based on the future weather data, the chilled water temperature data and the chilled water flow data; a first obtaining module configured to obtain cold station power data based on the demand cold quantity data and system efficiency data; and a second obtaining module configured to obtain the power adjustment space of the cold station based on the cold station power data.

[0016] Embodiments of the present application provide an electronic device, comprising: a memory, and one or more processors connected to the memory in communication; the memory has stored instructions executable by the one or more processors, and the instructions are executed by the one or more processors to cause the one or more processors to implement the steps of the method of any of the above embodiments.

[0017] Embodiments of the present application provide a computer-readable storage medium having stored thereon a computer program, the computer program being executed by a processor to implement the steps of the method of any of the above embodiments.

[0018] Embodiments of the present application provide a computer program product comprising a computer program, the computer program being executed by a processor to implement the steps of the method according to any of the above embodiments.

[0019] Through the power regulation space prediction method of the cold station provided by the present application, multi-element feature modeling and algorithm prediction can be combined, such as fusing multi-platform future weather data, chilled water temperature data, and reinforcement learning model demand cold quantity data, to realize dynamic evaluation of the power space of the cold station, and accurately quantify the upper and lower limits of the adjustable power of the cold station. Based on the boundary calculation of system efficiency data and cold station power data, the scheduling accuracy of the cold station power can be improved, the adjustable power interval of the cold machine and related auxiliary equipment in the cold station system is provided, and blind start or shutdown is avoided. At the same time, through the collaborative optimization of demand cold quantity data and cold station power extreme value, the energy saving potential is maximized, the unit load distribution and operation strategy are optimized on the premise of meeting the load demand, and the overall energy consumption of the cold station system is effectively reduced. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a cold station power regulation space prediction method provided by an embodiment of the present application is shown;

[0021] Figure 2 A flowchart of an example of a cold station power regulation space prediction method provided by an embodiment of the present application is shown;

[0022] Figure 3 A schematic diagram of a cold station power regulation space prediction device provided by an embodiment of the present application is shown;

[0023] Figure 4 A block diagram of an electronic device provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0024] Embodiments of the present application are described below in detail, examples of the embodiments are shown in the drawings, wherein the same or similar notations represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0025] With the increasing demand for air conditioning and refrigeration in large public buildings and industrial production, the proportion of chiller units and their auxiliary equipment (cooling towers, water pumps, etc.) in the total energy consumption of air conditioning systems continues to rise. The temperature, flow rate and other parameters of the chilled water side and the cooling water side directly determine the operating load and efficiency of the unit.

[0026] However, in actual engineering, the adjustment of the power of the cold station system is affected by many complex factors, including weather data (such as ambient temperature, humidity, air pressure, etc.), building load fluctuations (such as personnel density, equipment heat dissipation, and diurnal cycle changes), system efficiency of the cold station (such as cooling tower and water pump efficiency, chiller efficiency), and other unit-specific characteristics (such as part-load characteristics, start-stop strategies, etc.). The dynamic changes of these factors make the adjustment of the power of the cold station complex and varied, and the usual scheduling method is difficult to assess the adjustable power space of the cold station in real time and accurately, resulting in the inability to fully exploit the energy-saving potential.

[0027] In the usual operation scheduling, the operation and maintenance personnel mainly rely on existing experience or simple rules, such as "priority in advance" or "turning on and off in turn", to adjust the operation of the chiller and the auxiliary equipment. This extensive scheduling mode lacks prediction of the power adjustment space and is difficult to optimize the operation state of the unit in real time, often resulting in poor operating efficiency of the unit under partial load, and even over-regulation or under-regulation, affecting energy utilization efficiency. In addition, the usual scheduling method has low prediction accuracy of the power adjustment space and cannot comprehensively consider the mutual influence and dynamic characteristics of each component of the cold station system, resulting in insufficient precision and flexibility of power adjustment.

[0028] In view of the defects of the usual scheduling method of the cold station system, the present application provides a method for predicting the power adjustment space of a cold station. This method can combine multi-feature modeling and algorithm prediction to quantitatively evaluate the power adjustment space of the cold station.

[0029] Figure 1 A flowchart of a method for predicting the power adjustment space of a cold station provided by an embodiment of the present application.

[0030] As shown in Figure 1 The present application provides a method 100 for predicting the power adjustment space of a cold station, the method 100 comprising:

[0031] At step 110, future weather data, chilled water temperature data and chilled water flow data of the cold station are obtained.

[0032] Exemplarily, the obtained future weather data of the cold station comprises, for example, based on multi-platform future 24-hour every 15-minute weather forecast data, such as outdoor temperature, humidity, wind speed, solar radiation intensity. For multi-platform weather forecast data, Kalman filtering can be used to fuse multi-platform data to improve the reliability of the data. Specifically, the prediction variance of each platform for each prediction index data in the past 24 hours can be calculated:

[0033]

[0034] wherein, is the actual weather value such as temperature, humidity, etc. is the predicted weather value such as predicted temperature, humidity, etc. is the prediction variance; here N is generally 96.

[0035] Based on different indicators (outdoor temperature, humidity, wind speed, solar radiation intensity) of the prediction variance, different platform data is selected, i.e., the corresponding platform index data with low prediction variance is selected.

[0036] Exemplarily, the chilled water temperature data can include system design data. The weather forecast data is deeply fused with the cold station operation data to improve the cold load prediction accuracy.

[0037] At step 120, based on the future weather data, the chilled water temperature data and the chilled water flow data, the required cold quantity data of the cold station is predicted.

[0038] Exemplarily, the required cold quantity data of the cold station can be predicted by a reinforcement learning model. Specifically, under the condition of given future weather data and end device temperature of the cold station, for example, the future weather data is obtained based on step 110, and the end device temperature can be the full load on state of the end device, so as to predict the required cold quantity data.

[0039] At step 130, based on the required cold quantity data and the system efficiency data of the cold station, the cold station power data is obtained.

[0040] Exemplarily, the obtained cold station power data can include a cold station power maximum value and a cold station power minimum value.

[0041] At step 140, based on the cold station power data, the power regulation space of the cold station is obtained.

[0042] Exemplarily, the difference between the cold station power maximum value and the cold station power minimum value obtained in step 130 can be the power regulation space of the cold station.

[0043] The power regulation space prediction method of the cold station provided in the application can be combined with multi-element feature modeling and algorithm prediction, for example, fusing multi-platform future weather data, chilled water temperature data, and demand cold quantity data predicted by a reinforcement learning model, to realize dynamic evaluation of the power space of the cold station and accurately quantify the upper and lower limits of the adjustable power of the cold station. Based on boundary calculation of system efficiency data and cold station power data, the scheduling accuracy of the cold station power can be improved, the adjustable power interval of the cold machine and related auxiliary equipment in the cold station system is provided, and blind start or shutdown is avoided. At the same time, through the cooperative optimization of demand cold quantity data and cold station power extreme values, the energy saving potential is maximized, the load distribution and operation strategy of the unit are optimized on the premise of meeting the load demand, and the overall energy consumption of the cold station system is effectively reduced.

[0044] In another embodiment of the application, the demand cold quantity data includes a demand cold quantity maximum value and a demand cold quantity minimum value; the cold station power data includes a cold station power maximum value and a cold station power minimum value; and the cold station power data is obtained based on the demand cold quantity data and system efficiency data of the cold station, including: obtaining the system efficiency data based on future weather data and the chilled water temperature data; obtaining the cold station power maximum value based on the demand cold quantity maximum value and the system efficiency data; and obtaining the cold station power minimum value based on the demand cold quantity minimum value and the system efficiency data.

[0045] Illustratively, the demand cold quantity data includes two thresholds, a demand cold quantity maximum value and a demand cold quantity minimum value. Given the future weather data, and in the running state of the cold station, the demand cold quantity maximum value can make the chilled water temperature closest to the lowest chilled water return water temperature designed by the system, and when the return water temperature is close to and not lower than the lowest chilled water return water temperature, the indoor temperature can be maintained stable in the human comfort interval. The demand cold quantity minimum value can make the chilled water return water temperature closest to the highest chilled water return water temperature designed by the system, and when the return water temperature is close to and not higher than the lowest chilled water return water temperature, the indoor temperature can be maintained stable in the human acceptable interval. By distinguishing the demand cold quantity maximum value and the demand cold quantity minimum value, the power regulation demand under different comfort requirements can be accurately responded to.

[0046] Illustratively, in the process of obtaining the cold station power data based on the demand cold quantity data and the system efficiency data, since the system efficiency of the cold station is closely related to the cooling water temperature, the system efficiency data can be dynamically corrected by a neural network model. Through the dynamically corrected system efficiency, the influence of the cooling water temperature change on the system efficiency can be reflected, so that the demand cold quantity maximum value and the demand cold quantity minimum value can be combined to more accurately calculate the cold station power maximum value and the cold station power minimum value.

[0047] In the embodiments of the present application, by defining the maximum required cooling capacity and the minimum required cooling capacity, which respectively correspond to ensuring that the chilled water return temperature is close to the lower limit of the system design to maintain indoor comfort temperature and allowing the return water temperature to be close to the upper limit of the system design to meet basic refrigeration requirements, the cooling station cooling capacity adjustment range is quantified. In combination with the influence of cooling water temperature change on system efficiency, the neural network model is used to dynamically correct the efficiency parameter, and the maximum cooling station power and the minimum cooling station power are accurately calculated. Through the embodiments of the present application, the fine prediction of the power adjustment space of the cooling station can be realized, the unit operation strategy of the cooling station is optimized under the premise of ensuring user comfort, the energy efficiency is improved, and the flexible load response of the power grid is supported, thereby effectively solving the adjustment deviation and adjustment lag problems caused by the static estimation of system efficiency in the traditional method.

[0048] In another embodiment of the present application, the above method further comprises: obtaining system efficiency data of the cooling station; obtaining system efficiency data of the cooling station, comprising: obtaining historical required cooling capacity data and cooling water temperature data; based on the historical required cooling capacity data, querying the table to obtain initial system efficiency data, wherein the table comprises a plurality of system efficiency values corresponding to a plurality of cooling capacity gradient values; based on the future weather data and the cooling water temperature data, adjusting the initial system efficiency data to obtain the system efficiency data.

[0049] Exemplarily, the isolated forest algorithm can be used to remove abnormal data (such as sensor abnormalities and power surges) in historical data (such as historical required cooling capacity data). Based on the historical required cooling capacity data, the initial system efficiency data is obtained by querying the table, and the table is as follows:

[0050]

[0051] According to the historical required cooling capacity data and the system efficiency (COP), the highest system efficiency value corresponding to different cooling capacity in the past 365 days is found, and every 10% of the system design cooling capacity is taken as a gradient. corresponds to a highest The general required cooling capacity data will not be lower than 20% of the system design cooling capacity, and the cooling station system should be turned off for too low cooling capacity requirement, and fresh air refrigeration should be used.

[0052] After obtaining the initial system efficiency data through the above table, the system efficiency values corresponding to the maximum required cooling capacity and the minimum required cooling capacity are obtained by using the table lookup method, and then the ratio of the maximum required cooling capacity to the system efficiency value corresponding thereto is determined as the maximum cooling station power, and the ratio of the minimum required cooling capacity to the system efficiency value corresponding thereto is determined as the minimum cooling station power.

[0053] In another example, based on future weather data and cooling water temperature data, the initial system efficiency data is adjusted to obtain system efficiency data, comprising: inputting the future weather data into the cooling water temperature prediction model to obtain the cooling water temperature data; inputting the cooling water temperature data into the system efficiency prediction model to obtain the adjustment parameter; based on the adjustment parameter, the initial system efficiency data is adjusted to obtain the system efficiency data.

[0054] Exemplarily, the cooling water temperature prediction model can be a 3-layer fully connected neural network model: (outdoor temperature, humidity, wind speed, solar radiation), the cooling water temperature prediction model is trained based on historical data, indicating the correlation between future weather data and cooling water temperature data (cooling water return water temperature), and can predict the dynamically changing cooling water temperature data (Tcooling) in the next 24 hours. The system efficiency prediction model can be a 5-layer fully connected neural network model: The system efficiency prediction model can learn the non-linear mapping between the cooling water temperature data and the system efficiency data, and output the dynamically adjusted system efficiency data (Psystem). Finally, the adjusted system efficiency data under the given future weather data is obtained:

[0055]

[0056] In an embodiment of the present application, the cooling water return water temperature data in the next 24 hours is dynamically predicted by the cooling water temperature prediction model, and the mapping relationship between the future weather data and the cooling water temperature data is established; the system efficiency prediction model learns the dynamic influence law of the cooling water temperature data on the system efficiency data based on the predicted cooling water temperature data, thereby generating the adjusted system efficiency data matched with the actual working condition. The training process driven by historical data ensures that the cooling water temperature prediction model and the system efficiency prediction model fully mine the dynamic change characteristics of weather data and cold station operation, avoid static deviation depending on existing experience or simple rules, and finally realize high-precision calculation of system power data based on the predicted cooling water temperature data and the adjusted system efficiency data, to provide reliable data support for the power regulation space of the cold station.

[0057] In another embodiment of the present application, based on the cold station power data, the power regulation space of the cold station is obtained, comprising: determining the difference between the maximum value of the cold station power and the minimum value of the cold station power as the power regulation space.

[0058] Exemplarily, the calculation formula of the maximum value of the cold station power is:

[0059]

[0060] The calculation formula of the minimum value of the cold station power is:

[0061]

[0062] wherein, is the maximum cold station power, is the maximum cold station power, is the maximum demand cold quantity, is the minimum demand cold quantity, is the system efficiency data.

[0063] The difference between the maximum cold station power and the minimum cold station power is determined as the power adjustment space :

[0064]

[0065] The power adjustment space of the cold station can be provided to a third-party electricity price platform, and the platform dynamically formulates and issues instructions of a target control power (such as peak shaving, valley filling or demand response) that adapts to current power grid demand based on real-time electricity price signals and ranges, guides the cold station to adjust the operating power within a safe adjustable range, and realizes electricity cost optimization and power grid collaborative regulation and control. The power feasible region is deduced based on the device physical constraints and the demand cold quantity, and the energy-saving potential is visually expressed.

[0066] In another embodiment of the present application, based on future weather data, chilled water temperature data and chilled water flow data, demand cold quantity data of the cold station is predicted, including: inputting the future weather data, the chilled water temperature data and the chilled water flow data into a cold quantity prediction model for prediction, and outputting the demand cold quantity data of the cold station; the cold quantity prediction model is obtained by training in the following manner: obtaining first sample data, wherein the first sample data includes weather sample data, chilled water supply temperature sample data, chilled water return temperature sample data and chilled water flow sample data; based on the first sample data, the cold quantity prediction model is trained.

[0067] In another embodiment, the cold quantity prediction model includes a reinforcement learning model; based on the first sample data, the cold quantity prediction model is trained, including: constructing a state space based on the first sample data; based on the state space, an action space is obtained, wherein the action space includes the predicted demand cold quantity data; based on a reward function, a reward parameter for the action space is determined, wherein the reward parameter is associated with the chilled water return temperature; based on the reward parameter, the model parameters of the reinforcement learning model are updated to obtain a trained reinforcement learning model.

[0068] Exemplarily, the input data of the reinforcement learning model includes: historical weather data: outdoor temperature , humidity , chilled water temperature data: chilled water supply temperature , chilled water return water temperature , chilled water flow rate F. The output data of the reinforcement learning model includes: maximum demand cooling capacity and minimum demand cooling capacity under the condition of given weather data. The reinforcement learning model adopts a Deep Deterministic Policy Gradient (DDPG) reinforcement learning algorithm, which is suitable for continuous cooling action space prediction and can learn the optimal demand cooling capacity data Q policy. The reinforcement learning model can also be used for a cold load prediction module, future weather data, chilled water temperature data and chilled water flow data, for example, are input into the cold load prediction module through the running data input module for prediction. The static configuration data of the cold load prediction module, for example, includes the design parameters of the cold station: design cooling capacity, design power, chilled water supply and return water temperature, cooling water supply and return water temperature, design chilled water minimum flow rate, design chilled water return water maximum and minimum value. The training history data of the cold load prediction module, for example, includes: outdoor temperature and humidity, cold station cooling capacity / power, system efficiency, chilled water / cooling water main flow rate, pressure, temperature, real-time power of equipment. For the above data, an isolation forest algorithm can also be used to remove abnormal data, such as sensor abnormalities, power surges, etc.

[0069] The specific algorithm includes state space, action space, reward function, environment model, etc. The state space contains the key factors affecting the chilled water return water temperature: weather data , chilled water temperature data , , F. The state is defined as:

[0070]

[0071] The action space is the cooling capacity Q provided by the cold station, which is a continuous value; the reward function is designed according to the chilled water return water temperature , two threshold values are introduced: the cold station design minimum chilled water return water temperature : when the chilled water return water temperature is close to and not lower than this value, the room is most comfortable. The cold station design maximum chilled water return water temperature : when the chilled water return water temperature is close to and not higher than this value, the room temperature is acceptable.

[0072] The setting of the two design temperature thresholds of the cold station design minimum chilled water return water temperature and the cold station design maximum chilled water return water temperature, the general building design document selects = 14℃, = 18℃, which can be adjusted according to actual needs. The cooling capacity step size Choose an appropriate step size (usually choose 1-10) to balance the prediction accuracy and computational efficiency. Reward function tuning: adjust the decay coefficient k and negative reward value through experiments to ensure that the model converges to a reasonable policy.

[0073] The definition of the reward function is:

[0074]

[0075] where k is the decay coefficient, used to control the speed of the reward decreasing with the increase of chilled water return temperature. The reward gradually decreases from 1 (most comfortable) to -1 (unacceptable).

[0076] A 3-layer fully connected neural network trained using historical data is used as an environment simulator to predict the chilled water return temperature: ; input of the environment model: state and action . Output: chilled water return temperature .

[0077] In the training process of the cooling capacity prediction model, the first sample data can be the operation record covering 365 days, which is used as the training basis of the cooling capacity prediction model, such as 365 days of outdoor temperature and humidity , chilled water system parameters, chilled water supply and return temperature, and chilled water flow and corresponding cooling capacity Q of the cooling station. Then build the environment model, train a neural network using historical data to predict the chilled water return temperature ; after training, the model is used as an environment simulator for reinforcement learning.

[0078] First, initialize the policy network (Actor): input state S, demand cooling capacity data Q corresponding to the current action, initialize the value network (Critic): input state S and action Q, output Q value (evaluate the expected reward parameter of action Q), and synchronize the initialization of the target network (Target Actor and Target Critic). Then collect experience, use the Actor network to generate cooling capacity Q, and add exploration noise to enhance exploration ability. Execute demand cooling capacity data Q in the environment simulator to get chilled water return temperature , reward parameter r and next state s'. And store the current state, demand cooling capacity data corresponding to the current action, current reward parameter and next state (s, Q, r, s') to the experience replay buffer. Finally, update the model parameters of the reinforcement learning model according to the time difference error method, randomly sample data from the experience replay buffer. Update the Critic network: calculate the time difference error (TD-error), the target value is:

[0079]

[0080] where γ is the discount factor, Q' and μ' are the outputs of the target network. Update the actor network: boost the expected reward parameters using policy gradient. Soft update the target network: slowly update the target network parameters.

[0081]

[0082] wherein, wherein τ is the soft update rate.

[0083] Repeat the experience collection and network update steps until the policy converges.

[0084] In the technical scheme provided in the present application, the learning rates of the actor network and the critic network are set to 0.0001 and 0.001 respectively, and the discount factor is set to γ = 0.99, the differential learning rate balances the stability of policy optimization and value evaluation. By controlling the noise amplitude, the exploration ability of the action space is enhanced in the early stage of training, and gradually converges to the optimal policy in the later stage. The actor network and the critic network are 3-layer fully connected networks, and the number of neurons in each layer is 128 or 256. The shallow network (such as 128 neuron layer) extracts the basic features of the input state (weather condition, chilled water system parameter), and the deep network (such as 256 neuron layer) captures the nonlinear correlation, and finally outputs high-precision required cooling data.

[0085] In the embodiment of the present application, the reinforcement learning model is trained by historical weather data, chilled water temperature data and chilled water flow data, and the demand cooling maximum value (the most comfortable demand cooling) and the demand cooling minimum value (the acceptable demand cooling) of the cold station are dynamically predicted by combining the environment simulator, so as to realize high-precision quantification of the cooling adjustment interval; based on the design parameters and weather data, the cold station operation strategy is adaptively optimized, the energy efficiency is maximized under the premise of ensuring indoor comfort, and the problems of large prediction deviation and slow response of the traditional static model are effectively solved.

[0086] In another embodiment of the present application, the chilled water return water temperature associated with the reward parameter is obtained by: inputting the weather sample data, the chilled water supply water temperature sample data and the chilled water flow sample data of the cold station into the environment simulator for prediction to obtain the chilled water return water temperature.

[0087] Exemplarily, an environment model is trained based on the weather sample data, the chilled water supply water temperature sample data and the chilled water flow sample data of the cold station, the model is used for predicting the chilled water return water temperature, and after the training is completed, the model is used as the environment simulator of the reinforcement learning model.

[0088] In another embodiment of the present application, the cooling water temperature prediction model comprises a multi-layer fully connected neural network, and the cooling water temperature prediction model is trained by: obtaining second sample data and a cooling water temperature label corresponding to the second sample data, wherein the second sample data comprises historical outdoor temperature, historical outdoor humidity, historical wind speed, and historical solar radiation; processing the cooling water temperature sample data by using the multi-layer fully connected neural network to obtain cooling water temperature prediction data; determining the error between the cooling water temperature prediction data and the cooling water temperature label by using a root mean square error method, and adjusting the model parameters of the cooling water temperature prediction model based on the error to obtain a trained cooling water temperature prediction model.

[0089] Exemplarily, the cooling water temperature data is obtained by using the cooling water temperature prediction model, and the cooling water temperature prediction model is established by using, for example, a 3-layer fully connected neural network, and comprises an input layer, a hidden layer, and an output layer. The input layer can receive 4-dimensional features of the second sample data, such as historical outdoor temperature, historical outdoor humidity, historical wind speed, and historical solar radiation. The hidden layer is, for example, 2 layers, and the number of neurons in each layer is, for example, 128 and 64 respectively, which are used to extract weather data features. The output layer is, for example, 1 neuron, which directly outputs the prediction result of the cooling water temperature.

[0090] Exemplarily, in the training phase of the cooling water temperature prediction model, the historical outdoor temperature, the historical outdoor humidity, the historical wind speed, and the historical solar radiation of the cooling station in the recent period (for example, the previous 365 days) can be obtained from the operation log of the cooling station as the basic training data, the cooling water temperature prediction model is trained, the basic training data is input into the fully connected layer of the cooling water temperature prediction model, and the cooling water temperature in the training phase is generated. The root mean square error can also be used to calculate the error between the second sample data and the cooling water temperature label corresponding to the second sample data.

[0091] In another embodiment of the present application, the system efficiency prediction model comprises a multi-layer fully connected neural network, and the system efficiency prediction model is trained by: obtaining third sample data and a system efficiency label corresponding to the third sample data, wherein the third sample data comprises historical cooling water temperature data; processing the third sample data by using the multi-layer fully connected neural network to obtain system efficiency prediction data; determining the error between the system efficiency prediction data and the system efficiency label by using a root mean square error method, and adjusting the model parameters of the system efficiency prediction model based on the error to obtain a trained system efficiency prediction model.

[0092] Exemplarily, the system efficiency prediction model is established, for example, by using a 5-layer fully connected neural network, including an input layer, a hidden layer and an output layer. The input layer can receive 1-dimensional features of the third sample data, for example, historical cooling water temperature data. The hidden layer is, for example, 4 layers, and the number of neurons in each layer is, for example, 64, 32, 16 and 8 respectively, for capturing the relationship between the cooling water temperature and the system efficiency. The output layer is, for example, 1 neuron, outputting the dynamically adjusted system efficiency data.

[0093] In an embodiment of the present application, the cooling water temperature prediction model can be trained by associating historical weather data with cooling water return water temperature, and can dynamically predict future cooling water temperature, replacing the traditional static design value and improving environmental adaptability. The system efficiency prediction model can establish a dynamic mapping relationship between the cooling water temperature and the system efficiency based on the prediction result of the cooling water temperature, quantify the influence of environmental changes on the energy efficiency of the cold station, and provide real-time correction parameters for power calculation of the cold station.

[0094] Through the cooperative work of the two models, a more accurate power regulation space of the cold station can be provided, the energy utilization rate can be significantly improved, the operation and maintenance cost can be reduced, and the self-adaptive ability of the cold station system to complex weather conditions can be enhanced.

[0095] Figure 2 An example of the flowchart of the method for predicting the power regulation space of the cold station provided by an embodiment of the present application is shown.

[0096] The present application provides an example, and the content is described in combination with Figure 2 to better illustrate the content of the technical solutions provided by the present application:

[0097] Taking a certain office building cold station as an example: inputting the historical operation data of the cold station in the past 365 days to train the reinforcement learning model, obtaining the maximum demand cooling capacity and the minimum demand cooling capacity. The next 15-minute weather data predicted by the Kalman filter fusion of the multi-platform data is obtained: temperature 34 degrees, humidity 43%, and the minimum demand cooling capacity 1200kW and the maximum demand cooling capacity 1500kW output by the reinforcement learning model in the next 15 minutes.

[0098] The calculated power regulation range of the cold station in the next 15 minutes is 300kW to 450kW, and the power regulation space of the cold station is 150kW. Thus, the technical effects that can be achieved are: the prediction error of the reinforcement learning model is ≤10%, and the calculation accuracy rate of the power regulation space of the cold station is ≥90%.

[0099] Users can adjust the power consumption according to the electricity price and comfort requirements, and can also freely respond to the demand side of the power grid to obtain rewards accordingly.

[0100] Figure 3A schematic diagram of a power regulation space prediction device of a cold station is provided in an embodiment of the present application.

[0101] An embodiment of the present application further provides a power regulation space prediction device 300 of a cold station, characterized in that the device 300 comprises:

[0102] The acquisition module 310 is configured to acquire future weather data, chilled water temperature data and chilled water flow data of the cold station.

[0103] The prediction module 320 is configured to predict demand cooling capacity data of the cold station based on the future weather data and the chilled water flow data.

[0104] The first obtaining module 330 is configured to obtain cold station power data based on the demand cooling capacity data and system efficiency data.

[0105] The second obtaining module 340 is configured to obtain a power regulation space of the cold station based on the cold station power data.

[0106] It can be understood that the specific description of the power control device 300 of the cold station can refer to the description of the power control method 100 of the cold station in the foregoing.

[0107] An embodiment of the present application provides an electronic device, comprising a memory and one or more processors in communication connection with the memory; the memory stores instructions executable by the one or more processors, and the instructions are executed by the one or more processors to enable the one or more processors to implement the steps of the method of any one of the above embodiments.

[0108] An embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the method in any one of the above embodiments.

[0109] An embodiment of the present application provides a computer program product, which comprises instructions, and the instructions are executed by a processor of a computer device to enable the computer device to perform the steps of the method in any one of the above embodiments.

[0110] Figure 4 A block diagram of an electronic device is provided in an embodiment of the present application.

[0111] An embodiment of the present application provides an electronic device, comprising a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the method in any one of the above embodiments.

[0112] As shown in FIG. 4, in order to facilitate understanding, an embodiment of the present application shows a specific electronic device 400. Figure 4

[0113] ​Electronic device 400 is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. Electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present disclosure described and / or claimed in this document.

[0114] As shown in Figure 4 Device 400 includes a computing unit 401 that can perform various appropriate actions and processes in accordance with computer programs stored in a read-only memory (ROM) 402 or computer programs loaded into a random access memory (RAM) 403 from a storage unit 408. Various programs and data required for the operation of electronic device 400 can also be stored in RAM 403. Computing unit 401, ROM 402, and RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to bus 404.

[0115] Various components in electronic device 400 are connected to I / O interface 405, including an input unit 406, such as a keyboard, a mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a magnetic disk, a magneto-optical disk, etc.; and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0116] The computing unit 401 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 401 performs various methods described above. For example, in some embodiments, any one or more of the above methods can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded onto the RAM 403 and executed by the computing unit 401, one or more steps of any one or more of the above methods can be performed. Alternatively, in other embodiments, the computing unit 401 can be configured to perform any one or more of the above methods by other any appropriate means, such as by means of firmware.

[0117] It should be noted that the logical and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing logical functions and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions, or a combination thereof. For purposes of this application, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable medium upon which the program is printed, as the program can be electronically captured, for example, via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a manner that can be later executed by a computer. In some embodiments, the computer-readable medium can be a machine-readable medium.

[0118] It should be understood that portions of the application can be implemented in hardware, software, firmware, or combinations thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, implementation can be with any or a combination of the following technologies, which are all well-known in the art: a discrete logic circuit having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0119] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are contained in at least one embodiment or example of the present application. In the present application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0120] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application.

[0121] In addition, the terms "first", "second", etc. used in the embodiments of the present application are only for the purpose of description and can not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated in the embodiments. Therefore, the features defined with "first", "second" and the like in the embodiments of the present application can be explicitly or implicitly indicated to include at least one of the features. In the description of the present application, the meaning of the word "plurality" is at least two or two or more, such as two, three, four, etc., unless otherwise specifically limited in the embodiments.

[0122] In the present application, unless otherwise explicitly specified or limited in the embodiments, the terms "mounting", "connecting", "connecting" and "fixing" and the like appearing in the embodiments should be interpreted broadly, for example, the connection can be fixed connection, or detachable connection, or integral, can be understood, or can be mechanical connection, electrical connection, etc. Of course, it can also be directly connected, or indirectly connected through an intermediate medium, or it can be the internal communication of two elements, or the interaction relationship of two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific implementation situation.

[0123] In the present application, unless otherwise explicitly specified or limited, the first feature is "on" or "under" the second feature. The first and second features can be in direct contact or indirectly contact through an intermediate medium. Moreover, the first feature "above", "above" and "above" the second feature can be directly above or obliquely above the first feature, or only indicate that the horizontal height of the first feature is higher than that of the second feature. The first feature "below", "below" and "below" the second feature can be directly below or obliquely below the first feature, or only indicate that the horizontal height of the first feature is less than that of the second feature.

Claims

1. A method of predicting a power regulation space of a cold station, characterized in that, The method comprises: obtaining future weather data, chilled water temperature data and chilled water flow data of the cold station; based on the future weather data, the chilled water temperature data and the chilled water flow data, predicting the demand cold quantity data of the cold station, comprising: inputting the future weather data, the chilled water temperature data and the chilled water flow data into a cold quantity prediction model for prediction, and outputting the demand cold quantity data of the cold station; obtaining historical demand cold quantity data and cooling water temperature data; based on the historical demand cold quantity data, querying a table to obtain initial system efficiency data, wherein the table comprises a plurality of system efficiency values corresponding to a plurality of cold quantity gradient values; based on the future weather data and the cooling water temperature data, adjusting the initial system efficiency data to obtain the system efficiency data; based on the demand cold quantity data and the system efficiency data of the cold station, obtaining cold station power data; based on the cold station power data, obtaining the power adjustment space of the cold station; wherein the cold quantity prediction model comprises a reinforcement learning model, and the reinforcement learning model is obtained by training in the following way: obtaining first sample data, wherein the first sample data comprises weather sample data, chilled water supply temperature sample data, chilled water return temperature sample data and chilled water flow sample data; constructing a state space based on the first sample data; based on the state space, predicting to obtain an action space, wherein the action space comprises predicted demand cold quantity data; determining a reward parameter for the action space based on a reward function, wherein the reward parameter is associated with the chilled water return temperature, and the reward function is defined as: ； wherein, represents the lowest chilled water return temperature of the cold station design, represents the highest chilled water return temperature of the cold station design, represents the chilled water return temperature, and k is a decay coefficient. based on the reward parameter, updating the model parameters of the reinforcement learning model to obtain the trained reinforcement learning model.

2. The method of claim 1, wherein, The demand cold quantity data comprises a demand cold quantity maximum value and a demand cold quantity minimum value; the cold station power data comprises a cold station power maximum value and a cold station power minimum value; based on the demand cold quantity data and the system efficiency data of the cold station, obtaining cold station power data, comprising: based on the demand cold quantity maximum value and the system efficiency data, obtaining the cold station power maximum value; based on the demand cold quantity minimum value and the system efficiency data, obtaining the cold station power minimum value; wherein, based on the cold station power data, obtaining the power adjustment space of the cold station comprises: determining the difference between the cold station power maximum value and the cold station power minimum value as the power adjustment space.

3. The method according to claim 1 or 2, characterized in that, based on the future weather data and the cooling water temperature data, adjusting the initial system efficiency data to obtain the system efficiency data, comprising: inputting the future weather data into a cooling water temperature prediction model to obtain the cooling water temperature data; inputting the cooling water temperature data into a system efficiency prediction model to obtain an adjustment parameter; based on the adjustment parameter, adjusting the initial system efficiency data to obtain the system efficiency data.

4. The method of claim 1, wherein, The method comprises: The current state, the demand cold quantity data corresponding to the current action, the current reward parameter, and the next state are cached; The model parameters of the reinforcement learning model are updated according to the time difference error mode by sampling from the cached data.

5. The method of claim 1, wherein, The chilled water return water temperature associated with the reward parameter is obtained by: The weather sample data, the chilled water supply water temperature sample data, and the chilled water flow sample data of the cold station are input into an environment simulator for prediction to obtain the chilled water return water temperature.

6. The method of claim 3, wherein, The cooling water temperature prediction model comprises a multi-layer fully connected neural network, and the cooling water temperature prediction model is obtained by training in the following manner: Second sample data and cooling water temperature labels corresponding to the second sample data are obtained, wherein the second sample data comprises historical outdoor temperature, historical outdoor humidity, historical wind speed, and historical solar radiation; The cooling water temperature sample data is processed by a multi-layer fully connected neural network to obtain cooling water temperature prediction data; The error between the cooling water temperature prediction data and the cooling water temperature labels is determined by a root mean square error method, and the model parameters of the cooling water temperature prediction model are adjusted based on the error to obtain a trained cooling water temperature prediction model.

7. The method of claim 3, wherein, The system efficiency prediction model comprises a multi-layer fully connected neural network, and the system efficiency prediction model is obtained by training in the following manner: Third sample data and system efficiency labels corresponding to the third sample data are obtained, wherein the third sample data comprises historical cooling water temperature data; The third sample data is processed by a multi-layer fully connected neural network to obtain system efficiency prediction data; The error between the system efficiency prediction data and the system efficiency labels is determined by a root mean square error method, and the model parameters of the system efficiency model are adjusted based on the error to obtain a trained system efficiency prediction model.

8. A power regulation space prediction device for a cold station, characterized by, The device is used to execute the method according to any one of claims 1-7, and the device comprises: An acquisition module is configured to acquire future weather data, chilled water temperature data, and chilled water flow data of the cold station; A prediction module is configured to predict demand cold quantity data of the cold station based on the future weather data, the chilled water temperature data, and the chilled water flow data; A first obtaining module is configured to obtain cold station power data based on the demand cold quantity data and system efficiency data; A second obtaining module is configured to obtain a power adjustment space of the cold station based on the cold station power data. 9.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to implement the method of any one of claims 1-7. The processor executes the computer program to implement the method of any one of claims 1-7.

Images