技术领域technical field
本发明涉及区域供冷站优化控制领域，特别涉及区域供冷中冷冻水流量和回水温度的预测，用于优化供冷系统运行中的冷冻水生产、运输和分配。The invention relates to the field of optimal control of district cooling stations, in particular to the prediction of frozen water flow and return water temperature in district cooling, which is used to optimize the production, transportation and distribution of frozen water in the operation of the cooling system.
背景技术Background technique
在区域供冷项目的运行过程中，制备的冷冻水量是基于用户的冷量需求。根据冷量需求，区域供冷系统可以建立控制策略，并由此设置相关设备运行参数。During the operation of the district cooling project, the amount of chilled water prepared is based on the user's cooling demand. According to the cooling demand, the district cooling system can establish a control strategy, and thus set the relevant equipment operating parameters.
在区域供冷项目中，核心设备是制冷设备，通常是冷机。通过消耗电力或热量(蒸汽、热水或烟气)，制冷设备可以通过制冷循环产生冷冻水。大型制冷系统，包括区域制冷系统，通常配备有储能装置，有时还利用管网的热惰性作为储能方式。制冷设备的性能和运行成本取决于设备的负荷率、外部条件和能源价格。提前了解用户的冷量需求，可以使工厂的操作人员对每个制冷设备以及储能设备的运行进行最佳安排，以使运行成本最小化。In a district cooling project, the core equipment is the refrigeration equipment, usually a chiller. By consuming electricity or heat (steam, hot water or flue gas), refrigeration equipment can produce chilled water through a refrigeration cycle. Large refrigeration systems, including district cooling systems, are often equipped with energy storage devices, sometimes using the thermal inertia of the pipe network as a means of energy storage. The performance and operating costs of refrigeration equipment depend on the load rate of the equipment, external conditions and energy prices. Knowing the user's cooling demand in advance can enable the plant's operators to make optimal arrangements for the operation of each refrigeration device and energy storage device to minimize operating costs.
因此，精确的冷量需求预测对区域制冷站的运行至关重要。同时，冷冻水是通过供冷管网输送的，因此供冷管网的反应总是滞后于运行参数的调整。总而言之，及时和精确的冷量需求预测是区域制冷站运行优化的前提条件。Therefore, accurate forecasting of cooling demand is crucial to the operation of district cooling stations. At the same time, chilled water is transported through the cooling pipe network, so the response of the cooling pipe network always lags behind the adjustment of operating parameters. All in all, timely and accurate forecasting of cooling demand is a prerequisite for optimal operation of district cooling stations.
在现有技术中，主要通过如下方法预测供冷管网中的冷量需求：In the prior art, the cooling demand in the cooling pipe network is mainly predicted by the following methods:
1.经验/手动方法1. Empirical/Manual Approach
-主要由操作人员使用，其基于经验曲线(室外温度和冷冻水温度之间的关系)，这种方法依赖于操作人员的经验，准确性往往不够。- Mainly used by operators, it is based on empirical curves (the relationship between outdoor temperature and chilled water temperature), this method relies on the experience of operators, and the accuracy is often not enough.
-另一种解决方法是直接重复前一天的需求曲线(或可根据天气条件进行校正)，这种“手动”方法也往往不够精确，这降低了制冷生产效率并影响了运营成本。- Another solution is to directly repeat the previous day's demand curve (or it can be corrected for weather conditions), this "manual" method is also often inaccurate, which reduces cooling production efficiency and affects operating costs.
2.统计/机器学习方法2. Statistical/Machine Learning Methods
这是目前绝大多数现有的商业解决方案。这些方法可以根据过去类似条件下的运行数据给出精确的结果，通常一个好的机器学习至少需要一个完整的供冷季的运行数据，而对于新建区域供冷项目通常无法提供这类完整的历史数据。即使对于已经运行3年以上的区域供冷站，由于不断有新客户的连接，冷量需求仍在不断变化，那么机器已经学到的东西就不能应用于新的工况。This is the vast majority of existing commercial solutions today. These methods can give accurate results based on past operating data under similar conditions, usually a good machine learning requires at least one complete cooling season operating data, and such a complete history is usually not available for new district cooling projects data. Even for a district cooling station that has been in operation for more than 3 years, the cooling demand is still changing due to the continuous connection of new customers, then what the machine has learned cannot be applied to the new working conditions.
3.物理模型方法3. Physical model method
该方法目前极少使用，其依靠建筑热性能数值模型来模拟冷量需求。该方法不被普及的主要原因在于其需要有关建筑围护结构、冷量分配设备和使用情况的详细信息，特别是有关制冷使用和用户行为的信息很难获得，而这些信息对冷量需求有很大影响。This method, which is rarely used today, relies on numerical models of building thermal performance to simulate cooling demand. The main reason for the lack of popularity of this method is that it requires detailed information about the building envelope, cooling distribution equipment, and usage. In particular, information about cooling usage and occupant behavior is difficult to obtain, and this information has an impact on cooling demand. big impact.
因此，现有技术中的冷量需求预测方法和工具不能提供及时和准确的结果。由此导致区域供冷系统在缺乏准确预测的情况下不能运行在最佳状态下。Therefore, cooling demand forecasting methods and tools in the prior art cannot provide timely and accurate results. As a result, district cooling systems cannot operate optimally without accurate forecasting.
发明内容Contents of the invention
值得注意的是，本发明的目的在于克服背景技术中存在的缺陷和问题。It should be noted that the purpose of the present invention is to overcome the defects and problems existing in the background art.
为了这个目的，根据本发明的一个方面，本发明提出一用于预测供冷管网中的冷量需求的方法，其包括：For this purpose, according to one aspect of the present invention, the present invention proposes a method for forecasting cooling demand in a cooling network, which includes:
-建立初始库，所述初始库包括一套建筑热性能数值模型，其包括多个模型，每个模型分别对应于供冷管网中的各个建筑，且每个数值模型包含有限数量的参数并适于计算相应建筑的冷量需求；- establish an initial library, the initial library includes a set of building thermal performance numerical models, which includes a plurality of models, each model corresponds to each building in the cooling network, and each numerical model contains a limited number of parameters and Suitable for calculating the cooling demand of the corresponding building;
-使用相应建筑的历史冷量需求测量值和历史天气数据作为边界条件，校准所述一套建筑热性能数值模型中每个模型中的所述参数，从而获得一套参数化数值模型；- calibrating said parameters in each model of said set of building thermal performance numerical models using historical cooling demand measurements and historical weather data of the corresponding building as boundary conditions, thereby obtaining a set of parameterized numerical models;
-在预设的训练期内对所述一套参数化数值模型进行模拟，其中，所述训练期截止于当前时间，以获得一组分别与所述一套参数化数值模型中各数值模型相对应的模拟历史冷量需求时间序列；-Simulating the set of parameterized numerical models within a preset training period, wherein the training period ends at the current time, so as to obtain a set of parameters corresponding to each numerical model in the set of parameterized numerical models Corresponding simulated historical cooling demand time series;
-获得分别归属于每个所述模拟历史冷量需求时间序列的一组固定权重，以使所述一组模拟历史冷量需求时间序列的加权总和与所述训练期间内测量到的冷量需求相匹配，其中该组固定权重通过统计方法获得；- Obtain a set of fixed weights respectively assigned to each of the simulated historical cooling demand time series, so that the weighted sum of the set of simulated historical cooling demand time series is consistent with the measured cooling capacity demand during the training period , where the set of fixed weights is obtained by statistical methods;
-在预设的预测期内对所述一套参数化模型进行模拟，其中所述预测期始于所述当前时间，以所述预测期内的天气预报数据为边界条件，从而得到一组分别与所述一套参数化模型中各数值模型相对应的预测冷量需求时间序列；以及-Simulating the set of parameterized models within a preset forecast period, wherein the forecast period starts at the current time, using the weather forecast data in the forecast period as boundary conditions, thereby obtaining a set of respectively a forecasted cooling demand time series corresponding to each numerical model in the set of parameterized models; and
-将获得的所述固定权重应用于该组预测冷量需求时间序列，以便通过该组预测冷量需求时间序列的加权总和，获得所述预测期内的冷量需求预测。- Applying the obtained fixed weights to the group of predicted cooling demand time series, so as to obtain the cooling capacity demand forecast in the forecast period through the weighted sum of the group of predicted cooling capacity demand time series.
根据本发明的方法结合了物理模型和统计方法。利用建筑物的物理热性能模型，可以在未见的天气条件下推断出冷量需求。此外，由于使用统计方法，还可以捕捉到用户行为的影响。因此，本发明大大减少了统计方法所需的90％以上的训练数据。通过使用根据本发明的新方法，仅凭短暂的历史的运行数据(例如一周)就可以实现良好的冷量需求预测。The method according to the invention combines physical models and statistical methods. Using a model of the building's physical thermal performance, cooling requirements can be extrapolated for unseen weather conditions. In addition, due to the use of statistical methods, the influence of user behavior can also be captured. Therefore, the present invention greatly reduces the training data required by statistical methods by more than 90%. By using the novel method according to the invention, good forecasting of cooling demand can be achieved with only a short history of operating data (eg one week).
可选地，所述冷量需求包括供冷管网的总流量和/或回水温度，以及其他表示冷量需求的物理量。Optionally, the cooling demand includes the total flow of the cooling pipe network and/or the return water temperature, and other physical quantities representing the cooling demand.
可选地，每个数值模型包括描述热传递具有所述有限数量参数的线性或非线性微分方程。Optionally, each numerical model comprises linear or non-linear differential equations describing heat transfer with said finite number of parameters.
可选地，所述有限数量的参数是建筑物的物理参数和/或建筑物使用情景。进一步优选地，所述建筑物的物理参数包括地板面积、墙壁面积、热惰性、玻璃面积、空气处理装置的传热系数、墙壁、窗户和地板的热阻、太阳能透射率，所述建筑物使用场景包括入住率、室内温度设定值、换气次数、照明、设备和人员的内部得热，以及从太阳辐射获得的热量。Optionally, said limited number of parameters are physical parameters of the building and/or building usage scenarios. Further preferably, the physical parameters of the building include floor area, wall area, thermal inertia, glass area, heat transfer coefficient of the air handling unit, thermal resistance of walls, windows and floors, solar transmittance, and the building uses Scenarios include occupancy, room temperature setpoints, air change rates, lighting, internal heat gain from equipment and people, and heat gain from solar radiation.
可选地，在校准所述一套建筑热性能数值模型的所述参数后，通过修改至少部分所述物理参数和所述建筑使用场景来扩展所述一套参数化数值模型从而得到一套扩展参数化数值模型，其中，利用在以校准参数为中心的定义分布内使用随机选择来修改物理参数。从而得到包含更多模型的模型库，以便涵盖供冷管网中更多建筑物的冷负荷特性，提高预测精度。Optionally, after calibrating the parameters of the set of building thermal performance numerical models, the set of parameterized numerical models is extended by modifying at least part of the physical parameters and the building usage scenarios to obtain a set of extended A parametric numerical model in which the physical parameters are modified using random selection within a defined distribution centered on the calibration parameter. Thus, a model library containing more models is obtained, so as to cover the cooling load characteristics of more buildings in the cooling pipeline network, and improve the prediction accuracy.
可选地，也能基于对连接到供冷管网中的建筑物采取的测量，将所述一套参数化数值模型扩展为一套扩展参数化数值模型。优选地，对连接到所述供冷管网中的建筑物所采取的测量是由安装在建筑物中的一组传感器在规定的时间内获得的，以对参数进行校准。有利地，该组传感器中，至少有一个是测量建筑物从所述供冷管网中获取的流量、供冷管网的供水温度、供冷管网的回水温度、建筑物内选定位置的室内温度、建筑物内选定位置的室内空气湿度以及入住率。Optionally, the set of parameterized numerical models can also be expanded into a set of extended parameterized numerical models based on measurements taken on buildings connected to the cooling network. Preferably, the measurements taken on the buildings connected to said cooling network are obtained by a set of sensors installed in the buildings within a specified time to calibrate the parameters. Advantageously, at least one of the group of sensors measures the flow that the building obtains from the cooling pipe network, the supply water temperature of the cooling pipe network, the return water temperature of the cooling pipe network, the selected location in the building indoor temperature, indoor air humidity at selected locations within the building, and occupancy rates.
可选地，所述统计方法是线性回归。Optionally, said statistical method is linear regression.
可选地，所述训练期为当前时间之前的3至10天。Optionally, the training period is 3 to 10 days before the current time.
可选地，所述预测期为当前时间之后的24至48小时。Optionally, the prediction period is 24 to 48 hours after the current time.
根据本发明的另一个方面，本发明还提出一种设备，包括被配置为执行存储在计算机可读介质上的指令的处理器，以执行上述方法。According to another aspect of the present invention, the present invention also proposes a device, including a processor configured to execute instructions stored on a computer-readable medium, so as to perform the above method.
本发明其他方面的特点和优点将在下面具体实施方式中讨论，本领域技术人员基于以下实施例能够清楚地知道本发明的内容，以及所获得的技术效果。The features and advantages of other aspects of the present invention will be discussed in the following specific embodiments, and those skilled in the art can clearly understand the contents of the present invention and the obtained technical effects based on the following examples.
附图说明Description of drawings
应理解的是，在本发明中，除明显矛盾或不兼容的情况外，全部特征、变形方式和/或具体实施例可以根据多种组合相结合。It should be understood that, in the present invention, all the features, variants and/or specific embodiments can be combined in various combinations except in the case of obvious contradiction or incompatibility.
通过阅读以下作为非限制性说明的具体实施例，并结合附图，本发明的其它特征和优点将显而易见，图中：Other features and advantages of the present invention will become apparent by reading the following non-limiting description of specific embodiments, in conjunction with the accompanying drawings, in which:
-图1是根据本发明的用于预测供冷管网中的冷量需求的方法的一个实施例的示意图。- Figure 1 is a schematic diagram of an embodiment of a method for predicting cooling demand in a cooling network according to the present invention.
具体实施方式Detailed ways
以下是根据本发明的示例性实施例。下文中的相关定义近用于描述示例性实施例，而不是为了限制本发明的范围。由于这里描述的实施例是示例性的，它们也能被扩展至涉及本发明功能、目的和/或结构的修改。The following are exemplary embodiments according to the present invention. The relevant definitions below are only used to describe the exemplary embodiments, but not to limit the scope of the present invention. Since the embodiments described here are exemplary, they can also be extended to modifications involving the function, purpose and/or structure of the present invention.
图1示出了用于预测供冷管网中的冷量需求的方法的一个实施例。该示例性方法可以包括如下步骤：Fig. 1 shows an embodiment of a method for predicting cooling demand in a cooling network. The exemplary method may include the steps of:
-步骤1(图中①)- Step 1 (① in the figure)
基于例如描述热(和湿度)传递的线性和非线性微分方程系统，创建一个包含一套建筑热性能数值模型M1的初始库。该套模型包括至少一个建筑热性能数值模型，每个模型对应接入供冷管网的一栋(或一类)典型建筑物，例如酒店、办公楼等。这些模型能及时计算其所对应建筑的冷量需求，例如冷冻水流量和回水温度，以及还可以计算建筑物不同位置的室内温度。An initial library containing a set of numerical models M1 for the thermal performance of buildings is created, based eg on a system of linear and nonlinear differential equations describing heat (and humidity) transfer. The set of models includes at least one building thermal performance numerical model, and each model corresponds to a typical building (or a type) connected to the cooling pipe network, such as hotels, office buildings, etc. These models can calculate the cooling demand of the corresponding building in time, such as chilled water flow and return water temperature, and can also calculate the indoor temperature at different locations of the building.
这些模型具有有限数量的参数，包括与建筑几何形状、围护结构和冷量分配设备相关的物理参数以及包括和典型天数的有限数量建筑使用场景。物理参数包括例如，地板面积、墙壁面积、热惰性、玻璃面积、空气处理机组的传热系数、墙壁、窗户和地板的热阻，太阳能透射率。建筑物使用场景包括入住率、室内温度设定值、换气次数、照明、设备和人员的内部得热，以及直接或间接从太阳辐射获得的热量。These models have a limited number of parameters, including physical parameters related to building geometry, envelope and cooling distribution equipment, and a limited number of building usage scenarios including and typical number of days. Physical parameters include, for example, floor area, wall area, thermal inertia, glass area, heat transfer coefficient of air handling units, thermal resistance of walls, windows and floors, solar transmittance. Building usage scenarios include occupancy, indoor temperature set points, air change rates, lighting, internal heat gain from equipment and people, and heat gain from direct or indirect solar radiation.
更具体地，模型包括带有上述参数的方程组，以及求解方程组的求解器，该方程组给定输入(边界条件)：室外温度、室外湿度、太阳位置和太阳辐射量、内部得热。方程组的输出是流量和回水温度。More specifically, the model includes a system of equations with the above parameters, and a solver that solves the system of equations given inputs (boundary conditions): outdoor temperature, outdoor humidity, sun position and solar radiation, internal heat gain. The output of the equation system is flow rate and return water temperature.
进一步地，模型的方程描述了建筑中的热量和质量(湿度)传递，其中包括来自建筑围护结构的热量传递，换气的热量获得或损失，来自照明、设备和人的内部得热，空气中水的冷凝过程中释放的潜热，以及来自太阳辐射的热量获得，包括从玻璃直接进入的光线和从建筑围护结构的热传递。不同的模型可以有不同的方程组。以下为一个模型示例性的方程组：Further, the model's equations describe heat and mass (humidity) transfer in the building, including heat transfer from the building envelope, heat gain or loss from ventilation, internal heat gain from lighting, equipment, and people, air Latent heat is released during the condensation of water, as well as heat gain from solar radiation, including direct incoming light from glass and heat transfer from the building envelope. Different models can have different sets of equations. The following is an exemplary set of equations for a model:
Qbuildings＝Qenvelop+Qair+Qinternal heat+Qlatent+QsolarQbuildings ＝Qenvelope +Qair +Qinternal heat +Qlatent +Qsolar
其中：in:
Qbuildings为该模型对应(典型)建筑物的冷量需求Qbuildings is the cooling demand of the model corresponding to (typical) buildings
Qenvelope为通过建筑围护结构的热传递Qenvelope is the heat transfer through the building envelope
Qair为室内外空气流通及通风的换热负荷Qair is the heat exchange load of indoor and outdoor air circulation and ventilation
Qinternal heat为内部得热，包括建筑物内部人员、照明和设备Qinternal heat is internal heat gain, including people, lighting and equipment inside the building
Qlatent为空气中水气的汽化潜热Qlatent is the latent heat of vaporization of water vapor in the air
Qsolar为太阳辐射热量Qsolar is the solar radiation heat
同时，根据冷冻水的流量或回水温度与冷量的典型关系曲线，通过冷量需求计算冷冻水的流量和回水温度。这些典型的关系曲线取决于区域冷却项目的实际运行策略。At the same time, according to the typical relationship curve of chilled water flow or return water temperature and cooling capacity, the chilled water flow rate and return water temperature are calculated through the cooling capacity demand. These typical relationships depend on the actual operating strategy of the district cooling project.
-步骤2(图中②)- Step 2 (② in the figure)
校准模型中的参数以适应其对应的单个建筑过去的历史测量数据(在本发明没有时间段限制，但最好至少有一个季节(3个月的数据))。其使用相应的历史天气数据(与测量数据相同的时期，数据来自离区域供冷项目最近的气象站)作为模型的边界条件，利用一个基于逐渐逼近模拟的优化器寻找合适的物理参数和使用场景的组合，使模型计算的冷冻水流量和回水温度与实际测量值最匹配。理想情况下，使用的模型和校准参数与要预测的项目的模型和校准数据相近，但不一定来自该项目。换言之，针对一个典型建筑的模型可以被应用于相类似的其他建筑，例如具有类似规模的酒店和办公类。这一步的结果获得一套校准过的参数化模型M2。这套校准的参数化模型M2中的模型数量通常为5到20个，每一个对应于一个(典型建筑)。The parameters in the model are calibrated to their corresponding individual building past historical measurement data (no time period limit in the present invention, but preferably at least one season (3 months of data)). It uses the corresponding historical weather data (same period as the measured data, and the data comes from the weather station closest to the district cooling project) as the boundary conditions of the model, and uses an optimizer based on asymptotic approximation simulation to find suitable physical parameters and usage scenarios The combination of the chilled water flow and return water temperature calculated by the model can best match the actual measured values. Ideally, the model and calibration parameters used are close to those of the item being predicted, but not necessarily derived from that item. In other words, a model for a typical building can be applied to other similar buildings, such as hotels and offices of a similar size. The result of this step is a set of calibrated parametric model M2. The number of models in the set of calibrated parametric models M2 is usually 5 to 20, one for each (typical building).
此外，作为示例，上述历史数据通过传感器测得(如图1所示)。传感器能获得：In addition, as an example, the above historical data is measured by a sensor (as shown in FIG. 1 ). The sensor can obtain:
-历史的总冷冻水流量，以及供水和回水温度；- historical total chilled water flow, and supply and return water temperatures;
-对部分建筑物的测量，这些建筑物可以其代表了与供冷管网相连的典型建筑物(但不一定实际连接)，以便获得这些建筑物中- Measurements of buildings representing typical buildings connected to the cooling network (but not necessarily actually connected) in order to obtain
о必要数据：冷冻水流量，供水和回水温度оNecessary data: chilled water flow, supply and return water temperature
о可选数据：室内温度、室内湿度оOptional data: indoor temperature, indoor humidity
此外还需要获得以下额外的必要输入数据：In addition, the following additional necessary input data needs to be obtained:
-历史天气数据，包括供冷管网和被测建筑物的时期和位置，包括：-Historical weather data including the period and location of the cooling network and measured buildings, including:
о必要数据：室外干球温度。o Necessary data: Outdoor dry bulb temperature.
о可选数据：云层或太阳辐射，室外空气湿度，风速和风向。о Optional data: cloud cover or solar radiation, outside air humidity, wind speed and direction.
о可选数据：接入建筑物的使用率预测(例如：酒店预订)。o Optional Data: Access to building occupancy forecasts (eg: hotel reservations).
-预测期内的天气预报。- Weather forecast for the forecast period.
-步骤3(可选的，图中③)- Step 3 (optional, ③ in the picture)
可选地，对于具有大量建筑的供冷管网(区域供冷)，还可以通过在以"参数化模型"的参数为中心的定义分布范围内使用随机选择来修改一些物理参数，以及修改建筑使用方案的选择(随机或系统地)，从而扩展"参数化模型"的集合。Optionally, for cooling networks with a large number of buildings (district cooling), it is also possible to modify some physical parameters by using random selection within a defined distribution centered on the parameter of the "parametric model", as well as modify the building A selection of scenarios (randomly or systematically) are used, thereby extending the set of "parametric models".
这一步的目的是扩大参数化模型的中的模型数量，以便涵盖区域供冷项目中建筑热性能的更多可能性。对于每个"校准参数化模型"，可以根据以参数值为中心的正态分布(如利用正态分布公式)中依照为每个参数预设标准偏差随机选择新的参数值。这样，用新的物理参数就能创建更多的参数数值化模型。同时，使用场景也能被修改，可以类似物理参数一样随机修改，也可以根据建筑运行策略系统地修改。The purpose of this step is to expand the number of models in the parametric model in order to cover more possibilities of building thermal performance in district cooling projects. For each "calibration parameterized model", it can be calculated according to the normal distribution centered on the parameter value (such as using the normal distribution formula ) randomly select new parameter values according to the standard deviation preset for each parameter. In this way, more parametric numerical models can be created with new physical parameters. At the same time, usage scenarios can also be modified, either randomly like physical parameters, or systematically according to building operation strategies.
这一步的结果是一套扩展参数化数值模型M3，其也包括步骤2中获得的一套经过校准的参数化数值模型M2。通常，这个大数据库中的参数化模型的数量约是步骤1中选择的典型建筑的10倍。所有这些参数化模型可以用它们的特征来命名并存储在数据库中，后续根据需要，也可以继续添加新的模型。然后，所有这些现有的参数化模型可以在接下来的步骤中从数据库中选择。The result of this step is a set of extended parameterized numerical models M3, which also includes a set of calibrated parameterized numerical models M2 obtained in step 2. Typically, the number of parametric models in this large database is about 10 times that of a typical building selected in step 1. All these parametric models can be named with their characteristics and stored in the database, and new models can continue to be added later as needed. All these existing parametric models can then be selected from the database in the next steps.
当然，需要指出的是，该步骤并非必须，尤其是在供冷管网中(典型)建筑物不多的情况下，校准参数化模型已经能很好地覆盖现有建筑物，而不需要进行扩大模型库规模。Of course, it should be pointed out that this step is not necessary, especially when there are not many (typical) buildings in the cooling network, the calibration parameterized model can already cover the existing buildings well, and there is no need to carry out Expand the size of the model library.
-步骤4(图中④)- Step 4 (④ in the figure)
在预设的历史时间段内对一套参数化数值模型M2或一套扩展参数化数值模型M3进行模拟，其中该历史时间段以当前时间为终点，称为“训练期”，取决于项目的情况，类如为5天到10天。在步骤3中"参数化模型"的所有物理参数和使用场景都被输入到步骤1的物理模型中。通过将供冷管网所处位置的历史天气数据和历史冷冻水供水温度作为边界条件，在训练期内对所有参数化模型进行模拟。这一步的结果是，在训练期内各个参数化模型所对应模拟的历史冷量需求(流量和回水温度)的时间序列集。Simulate a set of parameterized numerical model M2 or a set of extended parameterized numerical model M3 within a preset historical time period, where the historical time period ends at the current time, called the "training period", depending on the project In some cases, such as 5 days to 10 days. All physical parameters and usage scenarios of the "parameterized model" in step 3 are input into the physical model in step 1. All parametric models are simulated during the training period with historical weather data and historical chilled water supply temperatures at the locations where the cooling network is located as boundary conditions. The result of this step is a time series set of simulated historical cooling demand (flow rate and return water temperature) corresponding to each parameterized model during the training period.
-步骤5(图中⑤)- Step 5 (⑤ in the figure)
使用统计方法，为每个时间序列(例如流量时间序列)赋予一个固定的权重，以便所有时间序列的加权总和与训练期中的测量得到的冷量(例如冷冻水流量)相匹配(权重可以是0)。同样地，也可以给其他时间序列(例如每个回水温度)赋予一个固定的权重。这一步的结果例如是一组流量时间序列的权重，和/或一组回水温度时间序列的权重。Using a statistical approach, assign a fixed weight to each time series (e.g. flow time series) such that the weighted sum of all time series matches the measured cooling during the training period (e.g. chilled water flow) (weight can be 0 ). Similarly, other time series (such as each return water temperature) can also be given a fixed weight. The result of this step is, for example, a set of weights for flow time series, and/or a set of weights for return temperature time series.
举例来说，上述权重可以通过如下示例性的利用回归算法的模型和过程获得。For example, the above weights can be obtained through the following exemplary model and process using a regression algorithm.
|Yregression-Ymeasure|→0 |Yregression -Ymeasure |→0
Yregression为通过回归算法计算的供冷管网的流量Yregression is the flow rate of the cooling pipe network calculated by the regression algorithm
p为建筑物数量p is the number of buildings
β为通过回归算法计算出的每个建筑物的固定权重β is the fixed weight of each building calculated by the regression algorithm
Yi为第i栋建筑物中冷冻水的流量Yi is the chilled water flow in the i-th building
C为常数截距C is the constant intercept
Ymeasure为测量得到的供冷管网的流量Ymeasure is the measured flow of the cooling pipe network
在统计学中，线性回归是对因变量和一个或多个自变量(也被称为输出和输入、响应和特征)之间的关系进行建模的一种线性方法。鉴于响应和特征的历史数据集，线性回归模型假定响应和特征之间的关系是线性的。因此，该模型的算式为：In statistics, linear regression is a linear method of modeling the relationship between a dependent variable and one or more independent variables (also known as outputs and inputs, response and features). Given a historical dataset of responses and features, linear regression models assume that the relationship between responses and features is linear. Therefore, the formula for this model is:
Y＝Xβ+εY=Xβ+ε
其中in
在本实施例中，Y代表“训练期”内测得的流量或回水温度时间序列，X代表“训练期”内针对P栋楼模拟得到的流量或回水温度时间序列，β表示P栋楼的模拟值被赋予的权重，ε表示所有流量或回水温度模拟的加权总和与实测流量或回水温度之间的差值。β^是β的估计值，它是通过最小化“训练期”内Y_预测的预测流量或回水温度时间序列与Y实测流量或回水温度时间序列之差的平方和得出的，即表示为残差平方和(RSS)。最小化过程称为普通最小二乘法(OLS)，如下公式所示：In this embodiment, Y represents the time series of flow or return water temperature measured during the "training period", X represents the time series of flow or return water temperature simulated for P buildings during the "training period", and β represents the time series of P buildings The simulated values of F are given weights, and ε represents the difference between the weighted sum of all flow or return temperature simulations and the measured flow or return temperature. β^ is an estimate of β obtained by minimizing the sum of squares of the difference between thepredicted flow or return temperature time series of Y_prediction and the measured flow or return temperature time series of Y_ during the "training period", i.e. is the residual sum of squares (RSS). The minimization process is called Ordinary Least Squares (OLS) and is shown in the following formula:
这是一种线性最小二乘法，用于估计线性回归模型中的未知参数。ε测量Y_预测和Y之间的差异。This is a linear least squares method for estimating unknown parameters in a linear regression model. ε measures the difference betweenY_predict and Y.
当然，其他回归模型也能被运用在本发明中，用来获得各个权重。Of course, other regression models can also be used in the present invention to obtain various weights.
-步骤6(图中⑥)- Step 6 (⑥ in the figure)
在预设的“预测期”内对“参数化模型”进行模拟，其中预测期从当前时间开始起算。该模拟以供冷管网所在位置的预报天气数据为边界条件，其中该天气预报来自最近的气象站。此步骤的结果是对应于“预测期”所模拟的“参数化模型”集的未来冷却需求的时间序列集，即分别与参数化模型中各模型相对应的预测冷量需求时间序列。The Parametric Model is simulated for a preset "forecast period", starting at the current time. The simulation is bounded by forecast weather data at the location of the cooling network, which is obtained from the nearest weather station. The result of this step is a set of time series of future cooling demands corresponding to the set of Parametric Models simulated by the Prediction Period, that is, time series of predicted cooling demands corresponding to each model in the Parametric Model.
据统计，在β^和X_预测的基础上，通过下面的算式得到预测期内P建筑物的模拟数据，其中Y_预测为预测期内预测的冷量需求，即预测冷冻水流量和回水温度。According to statistics, on the basis of β^ and X_prediction , the simulated data of building P in the forecast period can be obtained through the following formula, where Y_forecast is the forecasted cooling capacity demand during the forecast period, that is, the predicted chilled water flow and return water temperature .
随着时间的推移，历史数据会发生相应的变化。因此，如果能基于一定频率(例如每个小时)重复上述过程，预测结果能够及时更新(例如每小时)。Historical data changes accordingly over time. Therefore, if the above process can be repeated based on a certain frequency (eg every hour), the prediction result can be updated in time (eg every hour).
此外，为了评估模型的有效性，可以利用均方误差(MSE)和平均绝对误差(MAE)，如下所示。这两个指标的值都是用来评估训练期内测量和预测的冷量需求(流量或回水温度)时间序列之间的差异，其误差总是正值，随着误差接近零而减少。Furthermore, to evaluate the effectiveness of the model, the mean squared error (MSE) and mean absolute error (MAE) can be utilized as shown below. The values of both metrics are used to evaluate the difference between the measured and predicted cooling demand (flow or return temperature) time series during the training period, and the error is always positive and decreases as the error approaches zero.
其中，n是训练期内的测量流量或回水温度时间序列的数量，yi和分别是第i个测量和预测的流量或回水温度。where n is the number of measured flow or return temperature time series during the training period,yi and are the i-th measured and predicted flow or return temperature, respectively.
-步骤7(图中⑦)- Step 7 (⑦ in the figure)
将步骤5获得的权重集应用于步骤6获得的时间序列集，通过加权总和以计算预测期内的总的冷量需求，例如预测流量和预测回水温度，通过以下算式获得：Apply the weight set obtained in step 5 to the time series set obtained in step 6, and calculate the total cooling demand in the forecast period through the weighted sum, such as the predicted flow rate and predicted return water temperature, obtained by the following formula:
其中，Y_预测为供冷管网的预测冷量需求。Among them,Y_forecast is the forecasted cooling capacity demand of the cooling pipe network.
通过上述示例性的步骤，根据本发明的方法能准确且及时地预测出供冷管网中未来一定时间内的冷量需求。Through the above exemplary steps, the method according to the present invention can accurately and timely predict the cooling demand in the cooling pipeline network in a certain period of time in the future.
此外，在本实施例中，步骤4到7对每个新的预测都是重复的。而步骤1到3可以仅执行一次，或以较低的频率重复，用新的模型完成"参数化模型"的集合，这取决于对应的供冷管网中的建筑和其用冷情况是否有实质性变动，例如增加或减少接入的建筑物或增加或减少用冷(酒店的不同时节入住率)。Also, in this embodiment, steps 4 to 7 are repeated for each new prediction. However, steps 1 to 3 can be performed only once, or repeated at a lower frequency, with a new model to complete the collection of "parameterized models", depending on whether the buildings in the corresponding cooling network and their cooling conditions have Substantial changes, such as increasing or decreasing access to buildings or increasing or decreasing cooling (seasonal occupancy of hotels).
优选地，还能在"参数化模型"集中删除不必要的模型，例如：Preferably, also remove unnecessary models in the "parameterized models" set, for example:
-用统计方法分析归属于不同模型的权重历史，删除未被充分使用的模型；和/或- Statistically analyze the history of weights assigned to different models, removing underused models; and/or
-通过分析时间序列的历史，用统计学的方法，删除多余的模型，比如其结果与其他模型过于接近。-By analyzing the history of the time series, use statistical methods to remove redundant models, such as the results of which are too close to other models.
如图1所示，根据本发明的方法可以应用于整个区域供冷系统中，该系统用传感器和区域供冷站的控制平台进行数据监测。如上所述根据本发明方法获得的冷量需求预测被发送到区域供冷站的控制和监控平台，这些预测结果将被转化为控制信号并传输给制冷设备，然后制冷设备的工作负荷将被调整以满足供冷管网中的未来制冷需求。同时，由于安装在供冷管网中的监测传感器，供冷管网的实时运行数据被送回根据本发明的计算模型，这可以保持训练以优化供冷站的工作条件。As shown in Fig. 1, the method according to the present invention can be applied to the entire district cooling system, and the system uses sensors and the control platform of the district cooling station for data monitoring. As mentioned above, the cold demand forecast obtained according to the method of the present invention is sent to the control and monitoring platform of the district cooling station, and these forecast results will be converted into control signals and transmitted to the refrigeration equipment, and then the workload of the refrigeration equipment will be adjusted To meet the future cooling demand in the cooling network. At the same time, due to the monitoring sensors installed in the cooling pipe network, the real-time operation data of the cooling pipe network is sent back to the calculation model according to the present invention, which can keep training to optimize the working conditions of the cooling station.
以下将给出一个更为具体的实施例以说明根据本法明的方法来预测供冷网络中冷量需求。该实施例涉及一个区域供冷站，其主要用户是办公大楼和部分商业建筑。该区域供冷站生产冷冻水并通过供冷管网供应给用户。A more specific example will be given below to illustrate the prediction of cooling demand in the cooling network according to the method of the present invention. This embodiment relates to a district cooling station whose main users are office buildings and some commercial buildings. The district cooling station produces chilled water and supplies it to users through the cooling pipe network.
首先，根据建筑的类型，建立物理模型。如上述步骤1所述，这些模型包括一套描述热量和质量(湿度)传递的线性和非线性方程。例如，外墙的传热是基于以下传热方程：First, according to the type of building, build a physical model. These models include a set of linear and nonlinear equations describing heat and mass (humidity) transfer, as described in Step 1 above. For example, heat transfer in exterior walls is based on the following heat transfer equation:
Q＝K*A*ΔTQ=K*A*ΔT
Q为通过外墙传递的热量(W)Q is the heat transfer through the external wall (W)
K为传热系数(W/(m2*K))K is the heat transfer coefficient (W/(m2 *K))
ΔT为室内外温差(℃)ΔT is the temperature difference between indoor and outdoor (℃)
其中，in,
K＝1/(Ri+Rwall+Ro)K＝1/(Ri +Rwall +Ro )
R为热阻(内外表面换热热阻，以及外墙导热热阻)(m2*K/W)R is thermal resistance (thermal resistance of internal and external surfaces, and thermal resistance of external wall) (m2 *K/W)
因此，Rwall是步骤1中提到的物理参数之一，它将在上述步骤2中被校准。Therefore, Rwall is one of the physical parameters mentioned in step 1, which will be calibrated in step 2 above.
同时，根据冷却功率与流量或回水温度之间的典型关系曲线，来计算冷量需求。这些典型的关系曲线取决于该区域供冷站的实际运行策略。At the same time, the cooling demand is calculated according to the typical relationship curve between cooling power and flow rate or return water temperature. These typical relationship curves depend on the actual operation strategy of the district cooling station.
接着，在本实施例中选择10个带有测量数据的典型建筑。基于这些测量数据，在步骤2中优化器求解合适的物理参数和使用场景的组合，创建一个带有“校准参数化模型”的数据库。每个“校准参数化模型”都有一个特殊的物理参数和使用场景的值。例如，对于一个被选中的典型建筑，其热阻为4m2*K/W，考虑到该制冷项目所在地区建筑的普遍值，可以把外墙热阻范围定义在2至10m2*K/W内，然后优化器将尝试所有参数和方案的不同组合，以达到最适合的流量或回水温度曲线来适应测量数据。最终，对于其中一个"校准的参数化模型"，求解得到5.16m2*K/W的值。Next, 10 typical buildings with measurement data are selected in this embodiment. Based on these measurements, in step 2 the optimizer solves for the appropriate combination of physical parameters and usage scenarios, creating a database with a "calibrated parametric model". Each "Calibration Parametric Model" has a specific physical parameter and value for the usage scenario. For example, for a selected typical building, its thermal resistance is 4m2 *K/W, considering the general value of buildings in the area where the refrigeration project is located, the thermal resistance of the external wall can be defined in the range of 2 to 10m2 *K/W Inside, the optimizer will then try different combinations of all parameters and scenarios to arrive at the most suitable flow or return temperature profile to fit the measured data. Finally, for one of the "calibrated parametric models", the solution yielded a value of 5.16m2 *K/W.
然后，可选地，扩展数据库，以涵盖更多建筑热性能的可能性。例如，将上一步取得的5.16m2*K/W的热阻作为中心参数值，在此基础上建立一个正态分布。然后在±30％的标准偏差下，随机选择一些新的参数值。最后，获得总共获得104个参数化模型。Then, optionally, expand the database to cover more possibilities for building thermal performance. For example, take the thermal resistance of 5.16m2 *K/W obtained in the previous step as the central parameter value, and establish a normal distribution on this basis. Then, with a standard deviation of ±30%, some new parameter values are randomly selected. Finally, a total of 104 parametric models were obtained.
随后，所有这些“参数化模型”被输入到步骤1中提到的物理模型中，根据从最近的气候站获得的历史气候数据，模拟训练期内的冷冻水流量和回水温度。训练期例如可以被设定为7天。Subsequently, all these "parameterized models" were fed into the physical model mentioned in step 1 to simulate the chilled water flow and return water temperature during the training period based on the historical climate data obtained from the nearest climate station. The training period can be set to 7 days, for example.
接着，结合模拟预测结果，每个“参数化模型”通过回归计算给出一个比例(权重)，并计算出未来24至48小时小时内的冷量需求(冷冻水流量和回水温度)，在本实施例中给出24小时预测的示例。由此，通过“参数化模型”的加权总和得出未来24小时内104个建筑物的总冷量需求预测。在该过程中，β的估计值β^通过如下示例性的算式和统计学方法获得：Then, combined with the simulation prediction results, each "parameterized model" gives a proportion (weight) through regression calculation, and calculates the cooling capacity demand (chilled water flow and return water temperature) in the next 24 to 48 hours. An example of a 24-hour forecast is given in this embodiment. From this, the total cooling demand forecast of 104 buildings in the next 24 hours is obtained through the weighted sum of the "parametric model". In this process, the estimated value β^ of β is obtained through the following exemplary calculation formula and statistical method:
使用普通最小二乘法(OLS)方法，通过最小化测量值和预测值(Y和X)之间的差值，就能得到β^。应用公式其中在本实施例中X_预测来自基于"参数化模型"的24小时天气预报，以获得Y_预测。β^ is obtained by minimizing the difference between the measured and predicted values (Y and X) using the Ordinary Least Squares (OLS) method. Apply the formula Where in this exampleX_forecast comes from a 24-hour weather forecast based on a "parameterized model" to obtainY_forecast .
与简单地将前一天的24小时数据作为下一个24小时的预测的传统方法相比，根据本发明的模型所得到指标值(均方误差和平均绝对误差)较少。在本实施例中，根据本发明的方法的平均偏差约为10.4％，而传统重复前一天需求曲线的方法的平均误差值为12.2％。Compared with the traditional method of simply using the 24-hour data of the previous day as the next 24-hour forecast, the index values (mean square error and mean absolute error) obtained by the model according to the present invention are less. In this embodiment, the average deviation of the method according to the present invention is about 10.4%, while the average error value of the traditional method of repeating the previous day's demand curve is 12.2%.
最后，可以每隔1小时重复训练和预测(步骤4至7)，从而可以不间断地预测未来24小时的冷量需求。Finally, training and forecasting (steps 4 to 7) can be repeated every 1 hour, allowing for uninterrupted forecasting of cooling demand for the next 24 hours.
相对于传统的冷量需求预测方法，根据本发明方法，能在使用少量训练数据的情况下提高了预测的精度和效率。Compared with the traditional cold capacity demand forecasting method, according to the method of the present invention, the prediction accuracy and efficiency can be improved under the condition of using a small amount of training data.
本领域技术人员掌握多种实施例及多种变形及改进。尤其是，需明确的是，除明显矛盾或不兼容的情况外，本发明所记载的特征、变形方式和/或具体实施例可以相互结合。所有这些实施例及变形及改进都属于本发明的保护范围。Those skilled in the art are familiar with various embodiments and various modifications and improvements. In particular, it should be clear that the features, variants and/or specific embodiments described in the present invention may be combined with each other except in cases of obvious contradiction or incompatibility. All these embodiments and modifications and improvements belong to the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210105548.7ACN116557992A (en) | 2022-01-28 | 2022-01-28 | A Method of Forecasting Cooling Demand in Cooling Network |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210105548.7ACN116557992A (en) | 2022-01-28 | 2022-01-28 | A Method of Forecasting Cooling Demand in Cooling Network |
| Publication Number | Publication Date |
|---|---|
| CN116557992Atrue CN116557992A (en) | 2023-08-08 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210105548.7APendingCN116557992A (en) | 2022-01-28 | 2022-01-28 | A Method of Forecasting Cooling Demand in Cooling Network |
| Country | Link |
|---|---|
| CN (1) | CN116557992A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AT527607A2 (en)* | 2024-04-09 | 2025-02-15 | Avl List Gmbh | Parameterization procedure for generating parameterization information |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AT527607A2 (en)* | 2024-04-09 | 2025-02-15 | Avl List Gmbh | Parameterization procedure for generating parameterization information |
| Publication | Publication Date | Title |
|---|---|---|
| CN114484557B (en) | A method for regulating heating load of buildings based on target energy consumption control | |
| CN112856572B (en) | Automatic operation and regulation method of heating system based on characteristic parameter identification | |
| US6134511A (en) | Method and apparatus for improving building energy simulations | |
| JP2011214794A (en) | Air conditioning system control device | |
| CN102043907B (en) | Real-time cold load determination method of air-conditioner room | |
| Lu et al. | Operational optimization of district heating system based on an integrated model in TRNSYS | |
| CN107120721A (en) | A kind of central heating dynamic gas candidate compensation method | |
| Sha et al. | Machine learning-based cooling load prediction and optimal control for mechanical ventilative cooling in high-rise buildings | |
| Tüysüz et al. | Calibrating the building energy model with the short term monitored data: A case study of a large-scale residential building | |
| CN113065190B (en) | Uncertainty-based residential building heat supply amount calculation method | |
| CN107025334A (en) | Central heating system heat user integrated dynamic model method for building up and device | |
| Song et al. | Framework on low-carbon retrofit of rural residential buildings in arid areas of northwest China: A case study of Turpan residential buildings | |
| CN104331737A (en) | Office building load prediction method based on particle swarm neural network | |
| CN113435031A (en) | Parameter identification method and system for first-order ETP model of house to which air conditioner belongs | |
| Shi et al. | Evaluation of energy saving potential of HVAC system by operation data with uncertainties | |
| CN110673489A (en) | A heat load identification method for commercial building rooms | |
| CN116557992A (en) | A Method of Forecasting Cooling Demand in Cooling Network | |
| Im et al. | Empirical validation of building energy modeling using flexible research platform | |
| CN118134227B (en) | Short-term building cold-hot load prediction method and system based on hybrid model | |
| Bagheri et al. | Coupling building thermal network and control system, the first step to smart buildings | |
| CN118363318A (en) | Semi-physical simulation control method and management platform for building air conditioning system based on PLM system | |
| Weng et al. | RNN-based forecasting of indoor temperature in a naturally ventilated residential building | |
| Wang et al. | A Novel RC-Mapping Model of Building Air Conditioning Systems Dedicated to Demand Response and Training Condition Analysis | |
| Oh et al. | Component model calibration using typical AHU data for improved prediction of daily heat source energy consumption | |
| CN104866693A (en) | Optimal stop time prediction model of floor-radiating heating system |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |