Ultra-short-term heat exchange station heat load prediction algorithm

Abstract

The invention discloses an ultra-short-term heat exchange station heat load prediction algorithm. The algorithm comprises the following steps: establishing a heat exchange station heat load function; collecting operation data of the heat exchange station; preprocessing the operation data of the heat exchange station; establishing model training set data for the data, and substituting the model training set data into an ultra-short-term heat exchange station thermal load prediction model by adopting a BP neural network algorithm; performing data verification on the algorithm trained by the training set; calculating a data change rate lambda, and calculating a load index predicted value according to the data change rate; and calculating an ultra-short-term load prediction value and the like. According to the method, the heat demand of the heat exchange station is calculated from the room temperature of the heat user, the relation between multiple variables and the heat demand is defined, meanwhile, the prediction period is shortened, prediction data can be used for system regulation and control, precise heat supply of the heat exchange station is achieved, energy consumption of a heat source field station is reduced, and the operation efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of centralized heating, and in particular relates to an ultra-short-term heat exchange station heat load prediction algorithm. Background Art

[0002] Balancing heat demand and supply is one way to achieve energy savings in heating systems. The key to achieving this balance lies in effectively predicting users' heat demands. Heat exchange stations, a crucial link between heat sources and users, face load fluctuations that reflect changes in the heat demand of the buildings they serve. Therefore, accurately predicting the load of heat exchange stations and adjusting them accordingly is an effective way to achieve both a balance and energy savings.

[0003] Currently, there are many prediction algorithms for the heat load of heat exchange stations. Most of the prediction algorithms used and researched are long-term prediction algorithms, and the general prediction period is 1 day. For example, based on today's heating data and weather conditions, the total heating supply for tomorrow's day is predicted. This algorithm can only be used for daily heating supply allocation of heat source stations, and cannot achieve real-time daily heating supply adjustment and control. At the same time, this prediction method uses the overall operating parameters of the heating system as calculation data, and the prediction is also the overall heating supply, which cannot guide the real-time operation of the heat exchange station.

[0004] Moreover, the current load forecasting algorithms all have certain defects: 1) The forecasting method uses the overall operating parameters of the heating system as the calculation data, and the predicted value is also the overall heating supply of the heating system in one day, but it cannot be decomposed into hourly or even minutely heating supply, and can only be used for heating operation allocation between heat source stations; 2) There is a disconnection and inverted relationship between the overall predicted heating supply data and the heat exchange station. The correct idea should be that the heating supply of the heat source production station is solved by the heat demand of the heat exchange station, rather than being determined by the heat source station itself; 3) The granularity is not fine enough and the accuracy is not high. The accuracy of the load forecasting algorithm depends on the calculation granularity, that is, the smaller the system, the easier it is to achieve accurate load forecasting. The load forecast results of each system are then superimposed hourly according to time to obtain accurate hourly load forecast data for the heat source plant.

[0005] In practice, the heat load of a heat exchange station is affected by the outdoor temperature, the heat supply of the heating system, the thermal inertia of the building, solar radiation, and indoor heat dissipation equipment (the heat dissipation of this part is too small to be ignored). For a heating system, the data that can be monitored by the heat exchange station include: the secondary side supply and return water temperature at different times, the secondary side supply and return water pressure difference at different times, the system heat supply at different times, meteorological data at different times, and indoor temperature at different times. Therefore, these monitoring data should be used for ultra-short-term load forecasting at the heat exchange station level. Summary of the Invention

[0006] The present invention is proposed to overcome the shortcomings of the prior art and aims to provide an ultra-short-term heat exchange station heat load prediction algorithm.

[0007] The present invention is achieved through the following technical solutions:

[0008] An ultra-short-term heat exchange station heat load prediction algorithm includes the following steps:

[0009] (I) Establishing an ultra-short-term heat exchange station heat load prediction model;

[0010] (II) Collecting heat exchange station operation data;

[0011] (III) preprocessing the heat exchange station operation data collected in step (II);

[0012] (IV) Dividing the data preprocessed in step (III) into a training set and a validation set;

[0013] (V) Using the BP neural network algorithm, the training set is substituted into the ultra-short-term heat exchange station heat load prediction model in step (I) to establish a clear functional relationship;

[0014] (VI) Substituting the validation set into the explicit functional relationship established in step (V) to obtain the heat load index q of the heating station and verify the accuracy of the algorithm in step (V);

[0015] (VII) calculating the data change rate λ and calculating the load index prediction value based on the data change rate;

[0016] (VIII) Calculate the ultra-short-term load forecast value.

[0017] In the above technical solution, the step (I) of establishing the ultra-short-term heat exchange station heat load prediction model specifically includes the following steps:

[0018] (I-i) Establishing the initial expression of the ultra-short-term heat exchange station heat load prediction model;

[0019] (I-II) The initial expression for establishing the ultra-short-term heat exchange station heat load prediction model is converted to obtain the ultra-short-term heat exchange station heat load prediction model.

[0020] In the above technical solution, the initial expression of the ultra-short-term heat exchange station heat load prediction model is:

[0021] Q=w1*x1+w2*x2+w3*x3+…+w8*x8+b

[0022] Where: Q is the heat load of the heat exchange station, in kW;

[0023] w1~w8 are the weight coefficients of x1~x8, dimensionless;

[0024] b is the bias term, dimensionless;

[0025] x1 is the heat released by the heat exchange station at time n-2, in kW;

[0026] x2 is the heat released by the heat exchange station at time n-1, in kW;

[0027] x3 is the theoretical solar horizontal radiation, obtained according to theoretical calculation;

[0028] x4 is the normalized UV intensity, obtained from meteorological data;

[0029] x5 is the heat supplied by the heating system at time n-2, in kW, obtained from the heat meter data;

[0030] x6 is the heat supplied by the heating system at time n-1, in kW, obtained from the heat meter data;

[0031] x7 is the outdoor temperature in °C, obtained from meteorological data;

[0032] x8 is the indoor control temperature, in °C.

[0033] In the above technical solution, for multiple users, the indoor control temperature x8 is the average of the indoor control temperatures of the multiple users; the normalized ultraviolet intensity x4 is obtained by performing a base 10 logarithm calculation on the ultraviolet intensity data; the heat released by the heat exchange station x1 at time n-2 and the heat released by the heat exchange station x2 at time n-1 are obtained using the heat release calculation formula of the heat exchange station;

[0034] The calculation formula for the heat released by the heat exchange station is:

[0035] x i =kFΔt

[0036] Where: x i is the heat released by the heat exchange station at a certain moment, in kW, i = 1, 2; k is the heat transfer coefficient of the heat exchange station building, which is a constant and in kW / m 2 ℃; F is the surface area of ​​the heat exchange station building, which is a constant and the unit is m 2 ; Δt is the temperature difference between indoor and outdoor, in °C;

[0037] The calculation formula for the indoor and outdoor temperature difference is:

[0038] Δt=t i -t wi

[0039] Where: t i is the indoor temperature of the building at time i, in °C; t wiThe outdoor dry-bulb temperature at time i is in °C; i = n-1, n-2.

[0040] In the above technical solution, the expression of the ultra-short-term heat exchange station heat load prediction model is:

[0041]

[0042] Where: q is the heat load index of the heating station, the unit is W / m 2 ; Q is the heat load of the heat exchange station, unit is kW; s is the heating area, unit is m 2 .

[0043] In the above technical solution, the heat exchange station operation data includes the secondary side supply and return water temperature, supply and return water pressure, instantaneous heating capacity, cumulative heating capacity, user indoor temperature and meteorological data in the heat exchange station; the heat exchange station operation data is collected at an interval of 20 minutes.

[0044] In the above technical solution, the preprocessing includes filtering and abnormal data removal; the filtering adopts first-order filtering; the training set is 85% of the data preprocessed in step (III); and the validation set is 15% of the data preprocessed in step (III).

[0045] In the above technical solution, the calculation formula of the data change rate λ is:

[0046] λ=(q n ^ -q n-1 ) / q n-1

[0047] Where: q n-1 The actual heating index at time n-1, in W / m 2 ;q n ^ The predicted heating index at time n, in W / m 2 .

[0048] In the above technical solution, when λ≥35%, the calculation formula of the load index prediction value is q ^^ =q n-1 *(1+35%);

[0049] When -35%>λ>35%, the calculation formula for the load index prediction value is q ^^ =q n ^ ;

[0050] When λ≤-35%, the calculation formula for the load index prediction value is q ^^ =q n-1 *(1-35%);

[0051] In the above formula: q n-1 The actual heat supply at time n-1, in W / m 2 .

[0052] In the above technical solution, the ultra-short-term load forecast value Q ^ The calculation formula is:

[0053] Q ^ =q ^^ *s / 1000

[0054] Where: Q ^ is the ultra-short-term load forecast value, in kW; q ^^ is the load index prediction value, the unit is W / m 2 ; s is the heating area, unit is m 2 .

[0055] The beneficial effects of the present invention are:

[0056] The present invention provides an ultra-short-term heat load prediction algorithm for heat exchange stations. Starting from the room temperature of heat users, the heat demand of the heat exchange station is calculated, the relationship between multiple variables and heat demand is clarified, and the prediction cycle is reduced. The prediction data can be used for system regulation to achieve precise heating of the heat exchange station, thereby reducing the energy consumption of the heat source station and improving operating efficiency.

[0057] The present invention is applied in the actual operation and maintenance process. Each heat exchange station performs short-term load forecasting and summarizes it into a heat source plant load forecast. The overall load forecast is applied as the operation and maintenance adjustment schedule of the heat source plant to control the production of thermal energy. A function of the short-term load forecast of the heat exchange station and the secondary side water supply temperature can be established. The predicted load is used to calculate the secondary side water supply temperature required by the system and perform quality control on the secondary pipe network. The prediction results of the ultra-short-term heat exchange station heat load prediction algorithm can be used to adjust the secondary side water supply temperature and the water supply temperature required by the primary network. At the same time, the predicted heat loads of the heat exchange station at the same time are added together and can be used for the overall regulation of the heat source system. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of the method of the present invention;

[0059] Figure 2 This is a filtering processing diagram of the heat data released by the heat exchange station at time n-2 in Example 1 of the present invention;

[0060] Figure 3 This is a filtering processing diagram of the heat data released by the heat exchange station at time n-1 in Example 1 of the present invention;

[0061] Figure 4is a filtering processing diagram of heat data supplied by the heating system at time n-2 in Example 1 of the present invention;

[0062] Figure 5 is a filtering processing diagram of heat data supplied by the heating system at time n-1 in Example 1 of the present invention;

[0063] Figure 6 1 is a diagram showing the filtering process of outdoor temperature data in Example 1 of the present invention;

[0064] Figure 7 This is a graph of building heat load index data after filtering in Example 1 of the present invention (the red circle indicates abnormal values);

[0065] Figure 8 This is a comparison chart of the hourly heat supply of the heat exchange station and the outdoor temperature in Example 1 of the present invention;

[0066] Figure 9 This is a diagram showing the calculation results of the model training set in Example 1 of the present invention;

[0067] Figure 10 The result q1 of the ultra-short-term low-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Comparison chart with the actual result q1 (the red circle indicates the abnormal value);

[0068] Figure 11 The result q1 of the ultra-short-term low-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Linear relationship graph with actual result q1;

[0069] Figure 12 The result q1 of the ultra-short-term low-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Relative error diagram with actual result q1;

[0070] Figure 13 The result q2 of the ultra-short-term high-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Comparison chart with the actual result q2 (the red circle indicates the outlier);

[0071] Figure 14 The result q2 of the ultra-short-term high-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Linear relationship graph with actual result q2;

[0072] Figure 15 The result q2 of the ultra-short-term high-zone heat exchange station prediction algorithm in Example 1 of the present invention is ^ Relative error diagram with actual result q2;

[0073] Figure 16 The heat supply of the heat exchange station in Example 1 of the present invention is [(t g +t h) / 2-T in ]Linear correlation analysis diagram;

[0074] Figure 17 is the predicted value of the water supply temperature of the heat exchange station in Example 1 of the present invention, t g ^ and the actual water supply temperature t g Comparison picture.

[0075] For ordinary technicians in this field, other relevant drawings can be obtained based on the above drawings without any creative work. DETAILED DESCRIPTION

[0076] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be further described below with reference to the accompanying drawings and through specific implementation methods.

[0077] Example 1

[0078] like Figure 1 As shown, an ultra-short-term heat exchange station heat load prediction algorithm includes the following steps:

[0079] (I) Establishing an ultra-short-term heat exchange station heat load prediction model, specifically including the following steps:

[0080] (I-i) Establishing the initial expression of the ultra-short-term heat exchange station heat load prediction model;

[0081] The initial expression of the ultra-short-term heat exchange station heat load prediction model is:

[0082] Q=f(x1,x2,x3,…,x8)=w1*x1+w2*x2+w3*x3+…+w8*x8+b…(1)

[0083] Where: Q is the heat load of the heat exchange station, in kW;

[0084] w1~w8 are the weight coefficients of x1~x8, dimensionless;

[0085] b is the bias term, dimensionless;

[0086] x1 is the heat released by the heat exchange station at time n-2, in kW;

[0087] x2 is the heat released by the heat exchange station at time n-1, in kW;

[0088] x3 is the theoretical solar horizontal radiation, obtained according to theoretical calculation;

[0089] x4 is the normalized UV intensity, obtained from meteorological data;

[0090] x5 is the heat supplied by the heating system at time n-2, in kW, obtained from the heat meter data;

[0091] x6 is the heat supplied by the heating system at time n-1, in kW, obtained from the heat meter data;

[0092] x7 is the outdoor temperature in °C, obtained from meteorological data;

[0093] x8 is the indoor control temperature, in °C;

[0094] For multiple users, the indoor control temperature x8 uses the average value of the indoor control temperatures of multiple users;

[0095] The normalized ultraviolet intensity x4 is obtained by performing a logarithmic calculation with base 10 on the ultraviolet intensity data;

[0096] The heat released by the heat exchange station at time n-2 x1 and the heat released by the heat exchange station at time n-1 x2 are obtained using the heat release calculation formula of the heat exchange station;

[0097] The calculation formula for the heat released by the heat exchange station is:

[0098] x i =kFΔt

[0099] Where: x i is the heat released by the heat exchange station at a certain moment, in kW, i = 1, 2; k is the heat transfer coefficient of the heat exchange station building, which is a constant and in kW / m 2 ℃; F is the surface area of ​​the heat exchange station building, which is a constant and the unit is m 2 ; Δt is the temperature difference between indoor and outdoor, in °C;

[0100] The calculation formula for the indoor and outdoor temperature difference is:

[0101] Δt=t i -t wi

[0102] Where: t i is the indoor temperature of the building at time i, in °C; t wi The outdoor dry-bulb temperature at time i is in °C; i = n-1, n-2;

[0103] Therefore, x1=kF(t n-2 -t wn-2 ); x2=kF(t n-1 -t wn-1 ); Since k and F are constants, k and F are combined with the weight coefficient, so x1 and x2 are expressed as: x1=w1*(t n-2 -t wn-2 ); x2=w2*(tn-1 -t wn-1 );

[0104] (I-II) In order to make the load forecasting model universal and portable, the initial expression for establishing the ultra-short-term heat exchange station heat load forecasting model is transformed to obtain the ultra-short-term heat exchange station heat load forecasting model;

[0105] The expression of the ultra-short-term heat exchange station heat load prediction model is:

[0106]

[0107] Where: q is the heat load index of the heating station, the unit is W / m 2 ; Q is the heat load of the heat exchange station, unit is kW; s is the heating area, unit is m 2 ;

[0108] (II) Collecting heat exchange station operation data;

[0109] The heat exchange station operation data includes the secondary side supply and return water temperature, supply and return water pressure, instantaneous heat supply, cumulative heat supply, user indoor temperature and meteorological data in the heat exchange station;

[0110] The heat exchange station operation data is collected at an interval of 20 minutes;

[0111] (III) preprocessing the heat exchange station operation data collected in step (II);

[0112] The preprocessing includes filtering and abnormal data removal;

[0113] Filtering the data can reduce interference with feature data and lower the proportion of interference signals in useful signals. At the same time, in the digitization process, such as physical object scanning, data filtering helps remove erroneous points, redundant points, and measurement noise, thereby better extracting physical feature data.

[0114] The filtering adopts first-order filtering, and the formula of the first-order filtering is:

[0115] U n ^ =U n-1 ^ *a+U n *(1-a)……(5)

[0116] Where: U n ^ is the filtered data at time n; U n-1 ^ is the filtered data at time n-1; a is the weight value, a=0.05; U n is the actual measurement value at time n;

[0117] The filtering processing diagram of the heat data released by the heat exchange station at time n-2 and time n-1 is as follows Figure 2 、 3 As shown; the filtering processing diagram of the heat data supplied by the heating system at time n-2 and time n-1 is as follows Figure 4 、 5 As shown; the filtering processing diagram of outdoor temperature data is as follows Figure 6 As shown; the filtered building heat load index data is as follows Figure 7 As shown; the hourly comparison of heat supply of heat exchange station and outdoor temperature is as follows Figure 8 As shown;

[0118] Through data filtering, it can be seen that the continuity of the filtered data is significantly improved and the discreteness of the data is significantly reduced. The filtered data will help improve the accuracy of the prediction algorithm;

[0119] At the same time, it is necessary to remove outliers in the data to improve the accuracy of the model, such as Figure 7 Data in the red circle.

[0120] (IV) Dividing the data preprocessed in step (III) into a training set and a validation set;

[0121] The training set is 85% of the data pre-processed in step (III);

[0122] The validation set is 15% of the data preprocessed in step (III);

[0123] (V) Using the BP neural network algorithm, the training set is substituted into the ultra-short-term heat exchange station heat load prediction model to establish a clear functional relationship q = f(x1, x2, x3, …, x8);

[0124] The BP neural network is set up as follows: 1 input layer, 10 hidden layers, 1 output layer, the activation function adopts the tansig function, the training set is 70%, the validation set is 15%, and the test set is 15%, the loss function adopts the mean square error, and the optimizer adopts the Levenberg-Marquardt algorithm;

[0125] (VI) Substitute the validation set into the explicit functional relationship q = f(x1, x2, x3, …, x8) established in step (V) to obtain q and verify the accuracy of the algorithm in step (V);

[0126] The calculation results of the model training set are shown in the figure Figure 9 As shown;

[0127] The data verification results are as follows Figures 10-15 ; Figures 10-12 For verification one, that is, training set data verification; Figures 13-15For verification two, i.e., high-area data verification, the high-area data were filtered and abnormal data were eliminated.

[0128] from Figure 10 It can be seen that the BP neural network predicts the low-heat index q ^ The trend and data of the measured thermal index q in the low-heating area are very close, with a small error;

[0129] Figure 11 Display the BP neural network's predicted heating index q for the low-heat area ^ The linear correlation between the two data and the measured heating bottom area thermal index q is very high;

[0130] Figure 12 Shows the predicted low-heating zone heat index q ^ The maximum relative error with the measured thermal index q is 23.67%; the probability of the relative error being 0-10% is 99.78%, the probability of the relative error being 10-20% is 0.18%, and the probability of the relative error being 20-30% is 0.04%;

[0131] Figure 13 It can be seen that the BP neural network prediction model of the low heating area is used to predict the thermal index of the high heating area. The data change trend is consistent and the error is small.

[0132] Figure 14 It can be seen that the linear correlation between the predicted values ​​of the indicators in the high-heating area using the BP neural network prediction model in the low-heating area and the measured thermal indicators in the high-heating area is high;

[0133] Figure 15 It can be seen that the maximum relative error between the thermal index prediction value of the BP neural network in the heating high area and the actual value of the thermal index in the heating high area is 46.65%; the probability of the relative error between 0% and 10% is 88.16%, the probability of the relative error between 10% and 20% is 11.46%, the probability of the relative error between 20% and 30% is 0.32%, the probability of the relative error between 30% and 40% is 0.04%, and the probability of the relative error between 40% and 50% is 0.02%, which meets the engineering accuracy requirements.

[0134] (VII) calculating the data change rate λ and calculating the load index prediction value based on the data change rate;

[0135] The calculation formula of the data change rate λ is:

[0136]

[0137] Where: q n-1 The actual heat load index of the heating station at time n-1, in W / m 2 ;q n-1It is obtained by the ratio of the average heating data from time n-1 to n in the heat meter of the heat exchange station to the heating area; q n ^ The predicted heat load index of the heating station at time n, in W / m 2 ;q n ^ The ultra-short-term heat exchange station heat load prediction model with a clear functional relationship is obtained by step (V);

[0138] When λ≥35%, the calculation formula for the load index prediction value is:

[0139] q ^^ =q n-1 *(1+35%)……(7)

[0140] When -35%>λ>35%, the calculation formula for the load index prediction value is:

[0141] q ^^ =q n ^ ……(8)

[0142] When λ≤-35%, the calculation formula for the load index prediction value is:

[0143] q ^^ =q n-1 *(1-35%)……(9)

[0144] In the above formula: q n-1 The actual heat supply at time n-1, in W / m 2 ;q ^^ is the predicted value of the heat exchange station load index, in W / m 2 ;

[0145] from Figure 10 and Figure 13 It can be seen that no matter in the low or high heating supply area, the predicted values ​​of the thermal index of the BP neural network have many data anomalies (suddenly increasing or decreasing). In order to make the predicted data consistent with the actual situation, the data change rate λ is used to process the upper and lower limits of the predicted data (Formulas (7) to (9)) to improve data continuity.

[0146] (VIII) Calculation of ultra-short-term load forecast value Q ^

[0147] The calculation formula of the ultra-short-term load forecast value Q^ is:

[0148] Q ^ =q ^^ *s / 1000

[0149] Where: Q^ is the ultra-short-term load forecast value, unit is kW; q ^^ is the load index prediction value, the unit is W / m 2 ; s is the heating area, unit is m 2 .

[0150] q ^ is the load index value of the heat exchange station predicted by BP neural network, W / m 2 ;

[0151] q ^^ With q ^ and q n-1 The correction value is made based on formula (7)-(9), mainly to ensure the continuity of data. The final load forecast value is Q ^ The data is calculated by the formula Q ^ =q ^^ *s / 1000 to get.

[0152] (IX) Application of Ultra-Short-Term Load Forecasting

[0153] Each heat exchange station conducts short-term load forecasts and aggregates them into a heat source plant load forecast. The overall load forecast is used as the operation and maintenance adjustment schedule for the heat source plant to control the production of thermal energy.

[0154] A function of the short-term load forecast of the heat exchange station and the secondary side water supply temperature can be established, and the predicted load can be used to calculate the secondary side water supply temperature required by the system to perform quality control on the secondary pipe network.

[0155] Take the relationship between load forecast and secondary side water supply temperature as an example:

[0156] According to GB / T13754-2008 "Method for Determining Heat Dissipation of Heating Radiators" and JGJ142-2012 "Technical Specifications for Radiant Heating and Cooling", the heat dissipation equations for radiators and floor heating terminals are:

[0157] Q=K r [(t g +t h ) / 2-T in ] n

[0158] Where: K r is the heat dissipation coefficient at the end of the radiator or floor heating, which is a constant; t g is the secondary side water supply temperature, in °C; t h is the secondary side return water temperature, in °C; T in is the indoor temperature in °C; n is an exponent, usually n for radiators is 1.25 to 1.3, and n for floor heating is 1;

[0159] The above equation can be used to solve the secondary side water supply temperature t in the next control cycle. g , the specific formula is as follows:

[0160] t g.n =2(Q / K r +T in )-t h.n-1 ……(10)

[0161] Where: t g.n is the water supply temperature on the secondary side from time n to n+1, in °C; Q is the predicted heat load value, in kW; Kr is the heat dissipation coefficient at the radiator or floor heating terminal, which is a constant; T in is the target value of room temperature control, in °C; t h.n-1 is the secondary side return water temperature at time n, in °C (based on the assumption that the return water temperature remains unchanged from time n to time n+1).

[0162] according to Figure 16 According to the actual operation data of the low-zone system of the Dongjun project, Q and [(t g +t h ) / 2-T in ] There is a linear function relationship. It can be seen that the above theoretical analysis and data can confirm each other. The formula tg.n=2(Q / Kr+Tin)-th.n-1 can be used to calculate the secondary side water supply temperature. The comparison chart of the predicted value tg^ of the heat exchange station water supply temperature and the actual water supply temperature tg is shown in the figure below. Figure 17 As shown, the maximum error is 22.99%, the 0-5% error probability is 93.64%, the 5-10% error probability is 6.22%, the 10-15% error probability is 0.06%, the 15-20% error probability is 0.03%, and the 20-25% error probability is 0.04%.

[0163] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0164] The applicant declares that the above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention fall within the scope of protection and disclosure of the present invention.

Claims

1. An ultra-short-term heat exchange station heat load prediction algorithm, characterized by: The following steps are involved: (I) Establishing an ultra-short-term heat exchange station heat load prediction model; (II) Collecting heat exchange station operation data; (III) preprocessing the heat exchange station operation data collected in step (II); (IV) Dividing the data preprocessed in step (III) into a training set and a validation set; (V) Using the BP neural network algorithm, the training set is substituted into the ultra-short-term heat exchange station heat load prediction model in step (I) to establish a clear functional relationship; (VI) Substituting the validation set into the explicit functional relationship established in step (V) to obtain the heat load index q of the heating station and verify the accuracy of the algorithm in step (V); (VII) calculating the data change rate λ and calculating the load index prediction value based on the data change rate; (VIII) Calculate the ultra-short-term load forecast value.

2. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1 is characterized by: The step (I) of establishing the ultra-short-term heat exchange station heat load prediction model specifically includes the following steps: (I-i) Establishing the initial expression of the ultra-short-term heat exchange station heat load prediction model; (I-II) The initial expression for establishing the ultra-short-term heat exchange station heat load prediction model is converted to obtain the ultra-short-term heat exchange station heat load prediction model.

3. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 2 is characterized by: The initial expression of the ultra-short-term heat exchange station heat load prediction model is: Q=w1*x1+w2*x2+w3*x3+…+w8*x8+b Where: Q is the heat load of the heat exchange station, in kW; w1~w8 are the weight coefficients of x1~x8, dimensionless; b is the bias term, dimensionless; x1 is the heat released by the heat exchange station at time n-2, in kW; x2 is the heat released by the heat exchange station at time n-1, in kW; x3 is the theoretical solar horizontal radiation, obtained according to theoretical calculation; x4 is the normalized UV intensity, obtained from meteorological data; x5 is the heat supplied by the heating system at time n-2, in kW, obtained from the heat meter data; x6 is the heat supplied by the heating system at time n-1, in kW, obtained from the heat meter data; x7 is the outdoor temperature in °C, obtained from meteorological data; x8 is the indoor control temperature, in °C.

4. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 3 is characterized by: For multiple users, the indoor control temperature x8 is the average of the indoor control temperatures of the multiple users; the normalized ultraviolet intensity x4 is obtained by performing a base 10 logarithm calculation on the ultraviolet intensity data; the heat released by the heat exchange station x1 at time n-2 and the heat released by the heat exchange station x2 at time n-1 are obtained using the heat release calculation formula of the heat exchange station; The calculation formula for the heat released by the heat exchange station is: x i =kFΔt Where: x i is the heat released by the heat exchange station at a certain moment, in kW, i = 1, 2; k is the heat transfer coefficient of the heat exchange station building, which is a constant and in kW / m 2 ℃; F is the surface area of ​​the heat exchange station building, which is a constant and the unit is m 2 ; Δt is the temperature difference between indoor and outdoor, in °C; The calculation formula for the indoor and outdoor temperature difference is: Δt=t i -t wi Where: t i is the indoor temperature of the building at time i, in °C; t wi The outdoor dry-bulb temperature at time i is in °C; i = n-1, n-2.

5. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 2, characterized in that: The expression of the ultra-short-term heat exchange station heat load prediction model is: Where: q is the heat load index of the heating station, the unit is W / m 2 ; Q is the heat load of the heat exchange station, unit is kW; s is the heating area, unit is m 2 .

6. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1 is characterized by: The heat exchange station operation data includes the secondary side supply and return water temperature, supply and return water pressure, instantaneous heat supply, cumulative heat supply, user indoor temperature and meteorological data in the heat exchange station; the heat exchange station operation data is collected at an interval of 20 minutes.

7. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1 is characterized by: The preprocessing includes filtering and abnormal data removal; the filtering adopts first-order filtering; the training set is 85% of the data preprocessed in step (III); the verification set is 15% of the data preprocessed in step (III).

8. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1 is characterized by: The calculation formula of the data change rate λ is: λ=(q n ^ -q n-1 ) / q n-1 Where: q n-1 The actual heating index at time n-1, in W / m 2 ;q n ^ is the predicted heat supply index at time n obtained by the ultra-short-term heat exchange station heat load prediction model, in W / m 2 .

9. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1, characterized in that: When λ≥35%, the calculation formula of the load index prediction value is q ^^ =q n-1 *(1+35%); When -35%>λ>35%, the calculation formula for the load index prediction value is q ^^ =q n ^ ; When λ≤-35%, the calculation formula for the load index prediction value is q ^^ =q n-1 *(1-35%); In the above formula: q n-1 The actual heat supply at time n-1, in W / m 2 .

10. The ultra-short-term heat exchange station heat load prediction algorithm according to claim 1, characterized in that: The calculation formula of the ultra-short-term load forecast value Q^ is: Q ^ =q ^^ *s / 1000 Where: Q ^ is the ultra-short-term load forecast value, in kW; q ^^ is the load index prediction value, the unit is W / m 2 ; s is the heating area, unit is m 2 .

Images