CN109270842A  A kind of district heating model predictive control system and method based on Bayesian network  Google Patents
A kind of district heating model predictive control system and method based on Bayesian network Download PDFInfo
 Publication number
 CN109270842A CN109270842A CN201811251954.4A CN201811251954A CN109270842A CN 109270842 A CN109270842 A CN 109270842A CN 201811251954 A CN201811251954 A CN 201811251954A CN 109270842 A CN109270842 A CN 109270842A
 Authority
 CN
 China
 Prior art keywords
 bayesian network
 data
 network
 building
 source
 Prior art date
 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
 Granted
Links
 238000010438 heat treatment Methods 0.000 title claims abstract description 42
 230000001105 regulatory Effects 0.000 claims abstract description 18
 230000004044 response Effects 0.000 claims abstract description 16
 230000000875 corresponding Effects 0.000 claims description 9
 230000001537 neural Effects 0.000 claims description 8
 239000000446 fuel Substances 0.000 claims description 6
 230000001953 sensory Effects 0.000 claims description 5
 235000003140 Panax quinquefolius Nutrition 0.000 claims description 4
 235000005035 ginseng Nutrition 0.000 claims description 4
 235000008434 ginseng Nutrition 0.000 claims description 4
 238000009795 derivation Methods 0.000 claims description 3
 230000035832 Lag time Effects 0.000 claims description 2
 230000035648 Lagtime Effects 0.000 claims description 2
 240000002853 Nelumbo nucifera Species 0.000 claims 2
 235000006508 Nelumbo nucifera Nutrition 0.000 claims 2
 235000006510 Nelumbo pentapetala Nutrition 0.000 claims 2
 240000004678 Panax pseudoginseng Species 0.000 claims 1
 230000005611 electricity Effects 0.000 abstract description 4
 230000001808 coupling Effects 0.000 description 6
 238000010168 coupling process Methods 0.000 description 6
 238000005859 coupling reaction Methods 0.000 description 6
 238000000034 method Methods 0.000 description 6
 239000000203 mixture Substances 0.000 description 5
 238000005516 engineering process Methods 0.000 description 4
 241000208340 Araliaceae Species 0.000 description 3
 230000006399 behavior Effects 0.000 description 3
 230000005540 biological transmission Effects 0.000 description 2
 230000001276 controlling effect Effects 0.000 description 2
 230000004927 fusion Effects 0.000 description 2
 230000004048 modification Effects 0.000 description 2
 238000006011 modification reaction Methods 0.000 description 2
 235000015170 shellfish Nutrition 0.000 description 2
 241000196324 Embryophyta Species 0.000 description 1
 238000004458 analytical method Methods 0.000 description 1
 238000009529 body temperature measurement Methods 0.000 description 1
 230000000295 complement Effects 0.000 description 1
 238000010586 diagram Methods 0.000 description 1
 235000013399 edible fruits Nutrition 0.000 description 1
 238000010801 machine learning Methods 0.000 description 1
 239000011159 matrix material Substances 0.000 description 1
 230000000630 rising Effects 0.000 description 1
Classifications

 G—PHYSICS
 G05—CONTROLLING; REGULATING
 G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
 G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
 G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
 G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
 G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance

 G—PHYSICS
 G05—CONTROLLING; REGULATING
 G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
 G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
 G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
 G05B13/0265—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion
 G05B13/029—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric the criterion being a learning criterion using neural networks and expert systems

 G—PHYSICS
 G05—CONTROLLING; REGULATING
 G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
 G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
 G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
 G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
 G05B13/048—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators using a predictor
Abstract
The invention discloses a kind of district heating model predictive control system and method based on Bayesian network, method is the following steps are included: step S1, physical layer heat supply network data perception obtain in real time from source, net side and building side and update historical data；Step S2 constructs Bayesian network in conjunction with priori knowledge, and pass through Bayesian network forecasting thermal substation, the workload demand of building side according to historical data；Step S3, in conjunction with history data and real time data, obtains secondary side, primary side and source realtime control parameter by Bayesian Network Inference according to building workload demand；Step S4, according to history data and network topology, establish the time response curve that source is adjusted, net side is adjusted and building side is adjusted, determine source adjusting, net side valve and building side electricity tune valve regulation strategy, and according to this executive control operation, the hysteresis for eliminating regulating of heating net meets the realtime equilibrium of supply and demand, realizes the accurate heat supply on demand of heat user side.
Description
Technical field
It is one of the main foundation for realizing wisdom heat supply the invention belongs to the Dynamic matrix control field of heating system.Specifically relate to
And to a kind of district heating model predictive control system and method based on Bayesian network, source, thermal substation, net side are based on
The model that history large data sets are established, to realize the ondemand control accurate of the heating system equilibrium of supply and demand, heat user.
Background technique
Traditional central heating system has many characteristics, such as two big cores of close coupling, large time delay, thermal inertia and imbalance of supply and demand
Problem, i.e., by heating load fluctuation, indoor behaviour uncertainty, heat source side under tradition " sourcenetlotus " heat supply frame
Supplydemand mode caused by control measures unicity etc. is unbalance, and produce, transmission, loss, convert with " sourcenetlotus " thermal energy,
It consumes in each link, multilayer (heat source side, primary net, secondary network etc.) thermalhydraulic complicated coupling, thermal inertia, transmission lag
Property caused by supply and demand both ends regulation with respond on it is uncoordinated.Presently mainly pass through the warp of mechanism model combination operation personnel
Test and manual adjustment carried out to heat supply network, and due to the close coupling and hysteresis quality of regulating of heating net, operation personnel generally require experience adjust,
Stablize, the calibration process repeatedly that adjusts again, inefficiency and the technical level of operations staff is required very high.For regulating of heating net
Key problem, the present invention driven based on data model using big data and machine learning analytical technology, proposes one kind and be based on
The regional model load prediction control method and system of Bayesian network, since heat supply network load prediction has more impact factors, strong
The features such as coupling, uncertainty, Bayesian network model itself are a kind of uncertain correlation models, are not known with powerful
Property problem throughput, can effectively carry out multisource information expression and fusion.As a kind of uncertain inference based on probability
Method, Bayesian network to carry out heat supply network the characteristic of the more proper heat supply network load prediction of load prediction, and then eliminate heat supply network tune
Hysteresis is saved, realizes the accurate heat supply on demand of heat user side.
Summary of the invention
The object of the present invention is to provide a kind of district heating model predictive control system and method based on Bayesian network,
Based on history big data prediction source, thermal substation building side future shortterm thermic load and control source, thermal substation regulation ginseng
Number solves the problems, such as that heat supply network unbalanced supplydemand, regulation are inflexible, realizes the accurate heat supply on demand of heat user side.
The district heating model prediction based on Bayesian network that in order to solve the abovementioned technical problems, the present invention provides a kind of
Control method includes the following steps:
Step S1, physical layer heat supply network data perception obtain in real time from source, net side and building side and update historical data；
Step S2 carries out load prediction using Bayesian network method, according to historical data, constructs shellfish in conjunction with priori knowledge
This network of leaf, and pass through Bayesian network forecasting thermal substation, the workload demand of building side；The priori knowledge includes local spoke
According to, gas epidemic disaster, wind speed；
Step S3 is pushed away in conjunction with history data and real time data by Bayesian network according to building side workload demand
Reason obtains secondary side, primary side and source realtime control parameter；The control parameter of the secondary side, primary side and source includes
Circulation pump speed, valve opening, fuel quantity etc.；
Step S4 establishes source adjusting, net side adjusting and building side tune according to history data and network topology
The time response curve of section determines source adjusting, net side valve and building side valve regulated strategy, and executes control behaviour according to this
Make, eliminate the hysteresis of regulating of heating net, realizes the accurate heat supply on demand of heat user side.
In abovementioned technical proposal, step S1, physical layer heat supply network data perception obtains in real time from source, net side and building side
And update historical data, it may be assumed that be directed to spaceheating system, pass through its Internet of Things sensory perceptual system realtime data collection and more new historical number
According to the basis constructed as Bayesian network；Wherein the historical data includes: source load, thermal substation load, building side
Load, solar irradiation, gas epidemic disaster, thermal substation valve opening, circulation pump speed, building side valve opening.
Step S2, specific as follows: elder generation is according to historical data, bond area heating network operation situation and the passing experience of expert, really
It is fixed and rings the region heat supply network thermal substation, the variables set of building load, scope of a variable, known according to causality, that is, priori between variable
Know the prior Bayesian network structural model for constructing thermal substation and building side respectively；Such as the previous day load X_{1}, festivals or holidays X_{2}, season
X_{3}, temperature X_{4}, humidity X_{5}Etc. compositions influence load prediction variables set X={ X_{1}... ... X_{n}, then determine that each range of variables is become
Domain is measured, constructs the prior Bayesian network knot of thermal substation and building side respectively according to the causality (priori knowledge) between variable
Structure model.
Learnt and predicted by Bayesian network again: the history data set obtained by Internet of Things sensory perceptual system, input
Bayesian network structure and parameter are learnt in prior Bayesian network, obtain posterior Bayesian network, further according to pattra leaves
This neural network forecast obtains the load of thermal substation or building side；Wherein
History data set D={ the C of the input_{1}, C_{2}……C_{n}},C_{l}For a wherein data collection, i.e. one in database
Secondary record；
The bayesian network structure and parameter learning, comprising:
Bayesian network structure learning, according to history data set D, by Bayesian formulaSelection
Make p (S^{h} D) maximum network structure；
In above formula,
p(S^{h} D)  bayesian network structure is S at history data set D^{h}Probability；
S^{h} bayesian network structure；
D  history data set；
p(S^{h})  bayesian prior structure probability；
p(DS^{h})  structure likelihood；
P (D)  the probability of data set D does not influence structure；
By Bayesian network parameters prior distribution p (θ_{s}S^{h}, ε), by the history data set D of input, obtain Bayesian network
Posterior distrbutionp p (the θ of network parameter_{s}D,S^{h},ε)
In above formula,
S^{h} bayesian network structure；
The knowledge that ε  user has；
θ_{s} the parametric variable of prior probability；
p(θ_{s}S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is Bayes in the case where ε
The prior probability of parametric variable；
p(θ_{s}D,S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is ε, inputs history number
The posterior probability of Bayes's parametric variable in the case where according to collection D；
It is described by Bayesian network forecasting thermal substation, building side load, i.e.,
There is n example D={ C in history data set D_{1}, C_{2}……C_{n}, pass through Bayesian forecasting:
P(x_{N+1}D,S^{h})=∫ P (x_{N+1}θ_{s},D,S^{h})P(θ_{s}D,S^{h})dθ_{s}, choose P (x_{N+1}D,S^{h}) the corresponding x of maximum_{N+1}
As prediction result；
In above formula,
D  history data set；
S^{h} bayesian network structure；
θ_{s} the parametric variable of prior probability；
P(x_{N+1}D,S^{h})  in data set D, structure S^{h}In the case where event x_{N+1}The probability of generation.
Step S3, specific as follows:
Pass through pattra leaves in conjunction with history data and real time data according to building side workload demand obtained in step S2
This network reasoning obtains the control parameter of the control variable such as secondary side circulation pump speed, valve opening, in conjunction with transannular section parameter
And primary side, source constraint condition, derive the optimal RealTime Scheduling parameter of primary side, source；
It is described that the control ginseng that secondary side circulation pump speed, valve opening etc. control variable is obtained by Bayesian Network Inference
Number, i.e.,
Pass through Bayesian formulaProbability derivation is carried out with independency principle, infers and is going through
The secondary side circulation pump speed of corresponding maximum probability, valve opening etc. control variable under history load data, in this, as regulation
Amount；
In above formula,
A  control variables set, A={ A_{1}... ... A_{k}, k is control variable number
D '  history data set
P (A  D ')  the probability that control amount regulation parameter is A under history data set
The combination transannular section parameter and primary side, source constraint condition derive that primary side, source are optimal in real time
Scheduling parameter, it may be assumed that
By Bayesian inference, pass through Bayesian formulaIt is carried out with independency principle general
Rate derives, and primary side, source the control variable for inferring the corresponding maximum probability under historical load data include circulation pump speed
Degree, valve opening and fuel quantity, in this, as regulation amount.
In above formula,
A  the regulation amount of control variable, A={ A_{1}... ... A_{k}, k is control variable number
D "  history data set
P (A  D ")  the probability that control amount regulation parameter is A under history data set.
Step S4, specifically: according to history data, using neural network algorithm training source Load Regulation, net side
The time response model t=f (Δ Q, θ) of Load Regulation is changed by source obtained in step S2, thermal substation workload demand
Value Δ Q and the net side valve opening according to obtained in step S3, which are adjusted, obtains the slow of end response in θ input time response model
Stagnant time t, the time parameter method of control and regulation is determined with this, to eliminate the hysteresis of regulating of heating net, realizes that heat user side is ondemand
Accurate heat supply.
A kind of district heating model predictive control system based on Bayesian network of the invention, comprising:
Data acquisition module establishes building side data collection, thermal substation data set and net side for obtaining historical data respectively
Data set；
Load prediction module, for being driving with historical data, by Bayesian network to source, thermal substation and building side
Thermic load is predicted；
Parameter control module, source, thermal substation, building side thermic load according to prediction, is obtained by Bayesian Network Inference
To source, primary net, the realtime adjustment parameter of secondary network；
Time adjustment module, with historical data be driving, by neural network algorithm obtain load variations to end when
Between response model, with this determine adjusting strategy, eliminate regulating of heating net hysteresis, realize the accurate heat supply on demand of heat user side.
The present invention is compared to the advantage of existing method and technology:
1, independent of complicated modelling by mechanism, based on history big data, by the influencing characterisitic of different zones heat supply network
With the experience with heat source as priori knowledge, load prediction is carried out with datadriven by Bayesian network.
2, it is pushed away based on history big data by Bayesian network independent of the operation characteristic model of distinct device
Reason obtains equipment realtime monitoring parameter.
3, the time resolution characteristics of regulation are considered by neural network algorithm, and regulation implementation strategy is determined with this, it can be with
Eliminate regulating and controlling heat network hysteresis.
4, the present invention is not only limited to source and net side regulates and controls, but the thermal load demands based on building user side, to source
Side, net side and building side are regulated and controled simultaneously, reach control accurate, the equilibrium of supply and demand.
For the present invention based on Bayesian network, this network model of leaf itself is a kind of uncertain correlation model, is had
Powerful uncertain problem processing capacity can effectively carry out multisource information expression and fusion.As a kind of based on probability
Reasoning method under uncertainty selects Bayesian network not true to the heat supply network progress more proper heat supply network load prediction of load prediction
Fixed, close coupling, more influence factors characteristic, more accurate prediction result can be obtained.
Detailed description of the invention
Fig. 1 is the key step of the method for the present invention；
Fig. 2 is the building process of Bayesian network model；
Fig. 3 is the technology figure of the method for the present invention and system；
Fig. 4 is the prior Bayesian network structure chart in example.
Specific embodiment
The invention will now be described in further detail with reference to the accompanying drawings.Attached drawing is simplified schematic diagram, only with signal
Mode illustrates basic structure and process of the invention.
The invention belongs to the Model Predictive Control scopes of heating system.In conjunction with the priori knowledge or experience of heating network operation, lead to
Acquisition heat supply network historical load data collection is crossed, it is short by the source of Bayesian network forecasting heating system, thermal substation, building side future
Phase thermic load, then parameter data set is regulated and controled by history, source, thermal substation, building side regulation plan are obtained by Bayesian inference
Slightly.Heating system close coupling, thermal inertia, multiple constraint regulation problem are solved, realizes the accurate heat supply on demand of heat user side.
Make below in conjunction with method and system of the attached drawing to the district heating Model Predictive Control based on Bayesian network into one
The detailed description of step is provided based on bass neural network forecast thermic load, the detailed process of control parameter.
In conjunction with Fig. 1Fig. 3, the district heating model predictive control method of the invention based on Bayesian network is specifically included
Following steps:
Step S1, physical layer heat supply network data perception obtain in real time from source, net side and building side and update historical data；
Step S2 constructs Bayesian network in conjunction with the priori knowledges such as local irradiation, gas epidemic disaster, wind speed according to historical data
Network, and pass through Bayesian network forecasting thermal substation, the workload demand of building side；
Step S3 passes through Bayesian network in conjunction with history data and real time data according to end building workload demand
Reasoning obtains secondary side, primary side and source realtime control parameter；Secondary side, primary side and the source control parameter include
Circulation pump speed, valve opening, fuel quantity etc..
Step S4 establishes source adjusting, net side adjusting and building side tune according to history data and network topology
The time response curve of section determines source adjusting, net side valve and building side electricity tune valve regulation strategy, and executes control according to this
The hysteresis of regulating of heating net is eliminated in operation, realizes the accurate heat supply on demand of heat user side.
Wherein, step S1, physical layer heat supply network data perception obtain and more new historical in real time from source, net side and building side
Data；
In physical layer level, except the SCADA system using existing heating plant DCS system and heat supply network and heating network operation is obtained
Historical data, infull situation for data sets, need to do in terms of physical layer and arrange or improve: first part is to show
In model area primary side pipe network system, necessary temperature, pressuremeasuringpoint are supplemented in cell in the middle part of the thermal substation exit and heat supply network；Second
Part is made a concrete analysis of for secondary side demonstration quarter, drafts and supplements the representative indoor temperature measurement in representative building
The modification scheme of device；Part III is the electricity that building or building mouth unit are layed in for secondary side demonstration quarter complementary design
Dynamic hydraulic equilibrium adjusting device, and the modification scheme for degree of rising again, pressure, flow status measuring device is carried out in conjunction with heat death theory,
To support the accuracy controlling using data model support heat supply network under different working conditions.The historical data of acquisition includes source
Load, thermal substation load, building side load, solar irradiation, gas epidemic disaster, thermal substation valve opening, circulation pump speed, building
Side valve opening.The data set time frequency and data precision of acquisition will affect the accuracy of prediction regulation, therefore accurately number
According to the premise that perception is using big data driving and Bayesian network forecasting control heat supply network.
It is illustrated in figure 3 method and system structure chart of the invention, is spread wherein being supplemented in advance in secondary side exemplary cells
It is this hair set on building or the electrohydrodynamic balance regulation equipment of building mouth unit, differential manometer, flowmeter, the perception from building end
One of bright feature,
Step S2 constructs Bayesian network in conjunction with the priori knowledges such as local irradiation, gas epidemic disaster, wind speed according to historical data
Network, and pass through Bayesian network forecasting thermal substation, the workload demand of building side；
It is illustrated in figure 2 Bayesian network building process, variables set and scope of a variable are determined according to priori knowledge, such as the previous day
Load X_{1}, festivals or holidays X_{2}, season X_{3}, temperature X_{4}, humidity X_{5}Etc. compositions influence load prediction variables set X={ X_{1}... ... X_{n}, then
Determine that each range of variables obtains scope of a variable X_{1}∈ [a, b] ... ... X_{n}∈ [p, q], according to the causality between variable, (priori is known
Know) respectively construct thermal substation and building side prior Bayesian network structural model.Such as attached drawing 4, according to existing experience and because
The bayesian prior network of fruit relationship building.
Further the structure and parameter of Bayesian network is learnt by historical data.It is obtained by Internet of Things sensory perceptual system
The history data set obtained inputs in Bayesian network and learns to bayesian network structure and parameter, bayesian network structure
Study, according to history data set D, by Bayesian formulaSelection makes p (S^{h} D) maximum network knot
Structure.
In above formula,
p(S^{h} D)  bayesian network structure is S at history data set D^{h}Probability；
S^{h} bayesian network structure；
D  history data set；
p(S^{h})  bayesian prior structure probability；
p(DS^{h})  structure likelihood；
P (D)  the probability of data set D does not influence structure.
Bayesian network parameters study, by Bayesian network parameters prior distribution p (θ_{s}S^{h}, ε), pass through the history number of input
According to collection D, the Posterior distrbutionp p (θ of Bayesian network parameters is obtained_{s}D,S^{h},ε)
In above formula,
S^{h} bayesian network structure；
The knowledge that ε  user has；
θ_{s} the parametric variable of prior probability；
p(θ_{s}S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is Bayes in the case where ε
The prior probability of parametric variable；
p(θ_{s}D,S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is ε, inputs history number
The posterior probability of Bayes's parametric variable in the case where according to collection D.
The posteriority Bayes for being suitble to target area for thermal model is obtained according to historical data and Bayesian Network Learning training
Network predicts the following short term historical load data by posterior Bayesian network.
There is n example D={ C in history data set D_{1}, C_{2}……C_{n}, pass through Bayesian forecasting:
P(x_{N+1}D,S^{h})=∫ P (x_{N+1}θ_{s},D,S^{h})P(θ_{s}D,S^{h})dθ_{s}, choose P (x_{N+1}D,S^{h}) the corresponding x of maximum_{N+1}
As prediction result.
In above formula,
D  history data set；
S^{h} bayesian network structure；
θ_{s} the parametric variable of prior probability；
P(x_{N+1}D,S^{h})  in data set D, structure S^{h}In the case where event x_{N+1}The probability of generation.
Further, step S3 passes through shellfish in conjunction with history data and real time data according to end building workload demand
This network reasoning of leaf obtains secondary side, primary side and source realtime control parameter.
Pass through pattra leaves in conjunction with history data and real time data according to obtained building side workload demand response model
This network reasoning obtains secondary side control variable such as circulation pump speed, the control parameter of valve opening, in conjunction with transannular section parameter
And primary side, source constraint condition, derive the optimal RealTime Scheduling parameter of primary side, source.
Secondary side is wherein obtained by Bayesian Network Inference and controls variable such as circulation pump speed, the control ginseng of valve opening
Number, i.e.,
Pass through Bayesian formulaProbability derivation is carried out with independency principle, infers and is going through
The secondary side control variable of corresponding maximum probability includes circulation pump speed, valve opening, in this, as tune under history load data
Control amount.
In above formula,
A  control variables set, A={ A_{1}... ... A_{k}, k is control variable number
D '  history data set
P (A  D ')  the probability that control amount regulation parameter is A under history data set
The combination transannular section parameter and primary side, source constraint condition derive that primary side, source are optimal in real time
Scheduling parameter.I.e.
Equally by Bayesian inference, pass through Bayesian formulaWith independency principle into
Row probability derives, and primary side, source the control variable for inferring the corresponding maximum probability under historical load data include circulation
Pump speed, valve opening and fuel quantity, in this, as regulation amount.
In above formula,
A  the regulation amount of control variable, A={ A_{1}... ... A_{k}, k is control variable number
D "  history data set
P (A  D ")  the probability that control amount regulation parameter is A under history data set
Further, step S4 establishes source adjusting, net side is adjusted according to history data and network topology
Time response curve determines source adjusting, net side valve and building side electricity tune valve regulation strategy, and executes control behaviour according to this
Make, eliminate the hysteresis of regulating of heating net, realizes the accurate heat supply on demand of heat user side.I.e.
According to history data, using neural network algorithm training source Load Regulation, net side Load Regulation to end
Variation time response model t=f (Δ Q, θ), pass through obtained source, the value Δ Q and valve of the variation of thermal substation workload demand
The lag time t of end response is obtained in door aperture regulation θ input time response model, the time regulated and controled in advance is determined with this
T, to eliminate the hysteresis of regulating of heating net.
Taking the abovementioned ideal embodiment according to the present invention as inspiration, through the above description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention' entirely.The technology of this invention
Property range is not limited to the contents of the specification, it is necessary to which the technical scope thereof is determined according to the scope of the claim.
Claims (6)
1. a kind of district heating model predictive control method based on Bayesian network, which comprises the steps of:
Step S1, physical layer heat supply network data perception obtain in real time from source, net side and building side and update historical data；
Step S2 carries out load prediction using Bayesian network method, according to historical data, constructs Bayes in conjunction with priori knowledge
Network, and pass through Bayesian network forecasting thermal substation, the workload demand of building side；The priori knowledge include local irradiation,
Gas epidemic disaster, wind speed；
Step S3 is obtained in conjunction with history data and real time data by Bayesian Network Inference according to building side workload demand
To secondary side, primary side and source realtime control parameter；The control parameter of the secondary side, primary side and source includes circulation
Pump speed, valve opening, fuel quantity etc.；
Step S4 establishes what source adjusting, net side adjusting and building side were adjusted according to history data and network topology
Time response curve, determine source adjust, net side valve and building side valve regulated strategy, and according to this executive control operation,
The hysteresis of regulating of heating net is eliminated, realizes the accurate heat supply on demand of heat user side.
2. the district heating model predictive control method according to claim 1 based on Bayesian network, which is characterized in that
Step S1, physical layer heat supply network data perception obtain in real time from source, net side and building side and update historical data, it may be assumed that are directed to area
Domain heating system by its Internet of Things sensory perceptual system realtime data collection and updates historical data, as Bayesian network building
Basis；Wherein, the historical data include: source load, it is thermal substation load, building side load, solar irradiation, temperature, wet
Degree, thermal substation valve opening, circulation pump speed, building side valve opening.
3. the district heating model predictive control method according to claim 1 based on Bayesian network, which is characterized in that
Step S2, specific as follows:
First according to historical data, bond area heating network operation situation and the passing experience of expert, determining influences the region heat supply network heating power
It stands, the variables set of building load, scope of a variable, thermal substation and building is constructed according to causality, that is, priori knowledge between variable respectively
The prior Bayesian network structural model of space side；
The study and prediction of Bayesian network: the history data set obtained by Internet of Things sensory perceptual system inputs priori Bayesian network
Bayesian network structure and parameter are learnt in network, obtain posterior Bayesian network, is obtained further according to Bayesian network forecasting
To thermal substation or the load of building side；Wherein
History data set D={ the C of the input_{1}, C_{2}......C_{n}, C_{l}It is for a wherein data collection, i.e., primary in database
Record；
The bayesian network structure and parameter learning, comprising:
Bayesian network structure learning, according to history data set D, by Bayesian formulaSelection makes p
(S^{h} D) maximum network structure；
In above formula,
p(S^{h} D)  bayesian network structure is S at history data set D^{h}Probability；
S^{h} bayesian network structure；
D  history data set；
p(S^{h})  bayesian prior structure probability；
p(DS^{h})  structure likelihood；
P (D)  the probability of data set D does not influence structure；
By Bayesian network parameters prior distribution p (θ_{s}S^{h}, ε), by the history data set D of input, obtain Bayesian network ginseng
Several Posterior distrbutionp p (θ_{s} D, S^{h}, ε)
In above formula,
S^{h} bayesian network structure；
The knowledge that ε  user has；
θ_{s} the parametric variable of prior probability；
p(θ_{s}S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is Bayes's parameter in the case where ε
The prior probability of variable；
p(θ_{s} D, S^{h}, ε)  bayesian network structure be S^{h}, the state of knowledge that user has is ε, inputs history data set D
In the case where Bayes's parametric variable posterior probability；
It is described by Bayesian network forecasting thermal substation, building side load, i.e.,
There is n example D={ C in history data set D_{1}, C_{2}......C_{n}, pass through Bayesian forecasting:
P(x_{N+1} D, S^{h})=∫ P (x_{N+1}θ_{s}, D, S^{h})P(θ_{s} D, S^{h})dθ_{s}, choose P (x_{N+1} D, S^{h}) the corresponding x of maximum_{N+1}As pre
Survey result；
In above formula,
D  history data set；
S^{h} bayesian network structure；
θ_{s} the parametric variable of prior probability；
P(x_{N+1} D, S^{h})  in data set D, structure S^{h}In the case where event x_{N+1}The probability of generation.
4. the district heating model predictive control method according to claim 1 based on Bayesian network, which is characterized in that
Step S3, specific as follows:
The building side workload demand according to obtained in step 2 passes through Bayesian network in conjunction with history data and real time data
Reasoning obtains the control parameter of the control variable such as secondary side circulation pump speed, valve opening, in conjunction with transannular section parameter and one
The optimal RealTime Scheduling parameter of primary side, source is derived in secondary side, source constraint condition；
It is described that the control parameter that secondary side circulation pump speed, valve opening etc. control variable is obtained by Bayesian Network Inference,
I.e.
Pass through Bayesian formulaProbability derivation is carried out with independency principle, is inferred negative in history
The secondary side circulation pump speed of corresponding maximum probability, valve opening etc. control variable under lotus data, in this, as regulation amount；
In above formula,
A  control variables set, A={ A_{1}... ... A_{k}, k is control variable number
D '  history data set
P (A  D ')  the probability that control amount regulation parameter is A under history data set
The combination transannular section parameter and primary side, source constraint condition, derive the optimal RealTime Scheduling of primary side, source
Parameter, it may be assumed that
By Bayesian inference, pass through Bayesian formulaProbability is carried out with independency principle to push away
It leads, primary side, source the control variable for inferring the corresponding maximum probability under historical load data include circulation pump speed, valve
Door aperture and fuel quantity, in this, as regulation amount.
In above formula,
A  the regulation amount of control variable, A={ A_{1}... ... A_{k}, k is control variable number
D "  history data set
P (A  D ")  the probability that control amount regulation parameter is A under history data set.
5. the district heating model predictive control method according to claim 1 based on Bayesian network, which is characterized in that
Step S4 specifically:
According to history data, using the time response of neural network algorithm training source Load Regulation, net side Load Regulation
Model t=f (Δ Q, θ), by source obtained in step S2, the value Δ Q of thermal substation workload demand variation and according to step S3
Obtained in net side valve opening adjust the lag time t that end response is obtained in θ input time response model, determined with this
The time parameter method of control and regulation realizes the accurate heat supply on demand of heat user side to eliminate the hysteresis of regulating of heating net.
6. a kind of district heating model predictive control system based on Bayesian network characterized by comprising
Data acquisition module establishes building side data collection, thermal substation data set and net side data for obtaining historical data respectively
Collection；
Load prediction module, it is negative to source, thermal substation and building side heat by Bayesian network for being driving with historical data
Lotus is predicted；
Parameter control module, source, thermal substation, building side thermic load according to prediction, obtains source by Bayesian Network Inference
Side, primary net, the realtime adjustment parameter of secondary network；
Time adjustment module is driving with historical data, and it is loud to the time of end to obtain load variations by neural network algorithm
Model is answered, adjusting strategy is determined with this, regulating of heating net hysteresis is eliminated, realizes the accurate heat supply on demand of heat user side.
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

CN201811251954.4A CN109270842B (en)  20181025  20181025  Bayesian networkbased regional heat supply model prediction control system and method 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

CN201811251954.4A CN109270842B (en)  20181025  20181025  Bayesian networkbased regional heat supply model prediction control system and method 
Publications (2)
Publication Number  Publication Date 

CN109270842A true CN109270842A (en)  20190125 
CN109270842B CN109270842B (en)  20201113 
Family
ID=65194266
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

CN201811251954.4A Active CN109270842B (en)  20181025  20181025  Bayesian networkbased regional heat supply model prediction control system and method 
Country Status (1)
Country  Link 

CN (1)  CN109270842B (en) 
Cited By (9)
Publication number  Priority date  Publication date  Assignee  Title 

CN109935331A (en) *  20190402  20190625  北京航空航天大学  A kind of blood glucose prediction method and system based on multimodel dynamic comprehensive 
CN110136415A (en) *  20190419  20190816  中国农业科学院农业资源与农业区划研究所  A kind of temperature Monitoring in Cold Chain system and control method 
CN110287509A (en) *  20190425  20190927  浙江大学  Flexibility analysis and the fault of construction diagnosis of municipal heating systems and localization method and system 
CN110686377A (en) *  20190920  20200114  珠海格力电器股份有限公司  Control method for air conditioner temperature selfadaptive adjustment, computer readable storage medium and air conditioner 
CN110794679A (en) *  20191108  20200214  浙江大学  Prediction control method and system for load regulation of industrial steam supply system 
CN111473407A (en) *  20200410  20200731  浙江大学  Modelbased centralized heating system ondemand accurate regulation and control method 
CN111561732A (en) *  20200518  20200821  瑞纳智能设备股份有限公司  Heat exchange station heat supply adjusting method and system based on artificial intelligence 
CN111578371A (en) *  20200522  20200825  浙江大学  Datadriven accurate regulation and control method for urban centralized heating system 
CN111580382A (en) *  20200518  20200825  瑞纳智能设备股份有限公司  Unitlevel heat supply adjusting method and system based on artificial intelligence 
Citations (10)
Publication number  Priority date  Publication date  Assignee  Title 

CN101436057A (en) *  20081218  20090520  浙江大学  Numerical control machining tool heat error Bayes network compensation method 
WO2010020252A1 (en) *  20080819  20100225  Viking Genetics Fmba  Methods for determining a breeding value based on a plurality of genetic markers 
CN102521674A (en) *  20111207  20120627  福建省电力有限公司  Computer generating method for shortterm load forecasting in summer in power consumer electricityutilization information acquisition system 
US20140052606A1 (en) *  20120816  20140220  Infosys Limited  System and method for facilitating prediction of a loan recovery decision 
CN104008433A (en) *  20140603  20140827  国家电网公司  Method for predicting mediumandlongterm power loads on basis of Bayes dynamic model 
CN106897775A (en) *  20170125  20170627  浙江大学  Softmeasuring modeling method based on Bayes's integrated study 
CN107330621A (en) *  20170630  20171107  中国水利水电科学研究院  A kind of multireservoir outlet structure escape works evaluation method for failure based on Bayesian network 
CN107464017A (en) *  20170801  20171212  浙江大学  Based on the adaptive softsensor Forecasting Methodology with time difference Bayesian network 
CN107860057A (en) *  20170926  20180330  国网浙江省电力公司杭州供电公司  A kind of cogeneration of heat and power heating system thermic load economic optimization dispatching method 
CN108167035A (en) *  20171201  20180615  浙江大学  A kind of thermoelectricity cooperative system and method adjusted based on reheating accumulation of heat 

2018
 20181025 CN CN201811251954.4A patent/CN109270842B/en active Active
Patent Citations (10)
Publication number  Priority date  Publication date  Assignee  Title 

WO2010020252A1 (en) *  20080819  20100225  Viking Genetics Fmba  Methods for determining a breeding value based on a plurality of genetic markers 
CN101436057A (en) *  20081218  20090520  浙江大学  Numerical control machining tool heat error Bayes network compensation method 
CN102521674A (en) *  20111207  20120627  福建省电力有限公司  Computer generating method for shortterm load forecasting in summer in power consumer electricityutilization information acquisition system 
US20140052606A1 (en) *  20120816  20140220  Infosys Limited  System and method for facilitating prediction of a loan recovery decision 
CN104008433A (en) *  20140603  20140827  国家电网公司  Method for predicting mediumandlongterm power loads on basis of Bayes dynamic model 
CN106897775A (en) *  20170125  20170627  浙江大学  Softmeasuring modeling method based on Bayes's integrated study 
CN107330621A (en) *  20170630  20171107  中国水利水电科学研究院  A kind of multireservoir outlet structure escape works evaluation method for failure based on Bayesian network 
CN107464017A (en) *  20170801  20171212  浙江大学  Based on the adaptive softsensor Forecasting Methodology with time difference Bayesian network 
CN107860057A (en) *  20170926  20180330  国网浙江省电力公司杭州供电公司  A kind of cogeneration of heat and power heating system thermic load economic optimization dispatching method 
CN108167035A (en) *  20171201  20180615  浙江大学  A kind of thermoelectricity cooperative system and method adjusted based on reheating accumulation of heat 
NonPatent Citations (3)
Title 

YAN XIAOMING等: "Bayes discriminant analysis method for predicting the stability of open pit slope", 《 2011 INTERNATIONAL CONFERENCE ON ELECTRIC TECHNOLOGY AND CIVIL ENGINEERING (ICETCE)》 * 
仝营等: "基于机制模型与数据驱动的超临界锅炉性能在线预测方法", 《中国电机工程学报》 * 
郑惠丹等: "基于贝叶斯动态博弈理论的热价定价方法", 《山西建筑》 * 
Cited By (10)
Publication number  Priority date  Publication date  Assignee  Title 

CN109935331A (en) *  20190402  20190625  北京航空航天大学  A kind of blood glucose prediction method and system based on multimodel dynamic comprehensive 
CN110136415A (en) *  20190419  20190816  中国农业科学院农业资源与农业区划研究所  A kind of temperature Monitoring in Cold Chain system and control method 
CN110287509A (en) *  20190425  20190927  浙江大学  Flexibility analysis and the fault of construction diagnosis of municipal heating systems and localization method and system 
CN110686377A (en) *  20190920  20200114  珠海格力电器股份有限公司  Control method for air conditioner temperature selfadaptive adjustment, computer readable storage medium and air conditioner 
CN110794679A (en) *  20191108  20200214  浙江大学  Prediction control method and system for load regulation of industrial steam supply system 
CN111473407A (en) *  20200410  20200731  浙江大学  Modelbased centralized heating system ondemand accurate regulation and control method 
CN111473407B (en) *  20200410  20210330  浙江大学  Modelbased centralized heating system ondemand accurate regulation and control method 
CN111561732A (en) *  20200518  20200821  瑞纳智能设备股份有限公司  Heat exchange station heat supply adjusting method and system based on artificial intelligence 
CN111580382A (en) *  20200518  20200825  瑞纳智能设备股份有限公司  Unitlevel heat supply adjusting method and system based on artificial intelligence 
CN111578371A (en) *  20200522  20200825  浙江大学  Datadriven accurate regulation and control method for urban centralized heating system 
Also Published As
Publication number  Publication date 

CN109270842B (en)  20201113 
Similar Documents
Publication  Publication Date  Title 

CN109270842A (en)  A kind of district heating model predictive control system and method based on Bayesian network  
CN106958855B (en)  The hydraulically balanced model predictive control method of heating system and system  
CN108916986B (en)  Information physical fusion secondary pipe network variable flow hydraulic balance regulation and control method and system  
CN109253494B (en)  Control method of electric heat storage device based on heat load prediction  
CN106895478A (en)  A kind of heating network adjusting method and system  
Wojdyga  Predicting heat demand for a district heating systems  
Li et al.  An improved riskbenefit collaborative grey target decision model and its application in the decision making of load adjustment schemes  
CN103591637A (en)  Centralized heating secondary network operation adjustment method  
CN107726442A (en)  A kind of heat supply network balance regulation method  
CN108256672A (en)  A kind of central heating system the whole network balance adjusting method  
CN108876039B (en)  Electric energy quality prediction method for power distribution network with distributed power supply based on support vector machine  
CN107168101A (en)  Consider frequency modulation and the set speed adjustment system control parameters setting method of scleronomic constraint  
CN107909149A (en)  A kind of Temperature in Greenhouse Forecasting Methodology based on Genetic BP Neutral Network  
CN111473407B (en)  Modelbased centralized heating system ondemand accurate regulation and control method  
CN107766986A (en)  Leak integral form echo state network online study photovoltaic power Forecasting Methodology  
CN107868979A (en)  A kind of silicon single crystal diameter control method based on permanent casting speed control structure  
CN108579929A (en)  A kind of doublein and doubleout tube mill control system and control method based on RBF neural PREDICTIVE CONTROL  
CN109631150A (en)  A kind of accurate heat supply method of temperature based on secondary network, pressure and user's room temperature  
CN109242265A (en)  Based on the smallest Urban Water Demand combination forecasting method of error sum of squares  
CN108376294A (en)  A kind of heat load prediction method of energy supply feedback and meteorologic factor  
CN110866640A (en)  Power load prediction method based on deep neural network  
CN108303898B (en)  Intelligent scheduling method of novel solarair energy coupling cold and heat cogeneration system  
CN110474339A (en)  A kind of electric network reactiveload control method based on the prediction of depth generation load  
CN101635553A (en)  Integrated intelligent decoupling control method for generator excitation system and steam valve system  
CN107256436A (en)  The prediction and matching and control method of dissolving of thermal storage electric boiler and clean energy resource 
Legal Events
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  
GR01  Patent grant  
GR01  Patent grant 