You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+2-1
Original file line number
Diff line number
Diff line change
@@ -13,8 +13,9 @@ The EPANET Library is a pressurized pipe network hydraulic and water quality ana
13
13
14
14
Please see the [`version 2.1` Release Notes](https://github.com/OpenWaterAnalytics/EPANET/blob/master/ReleaseNotes2_1.md) for information relevant to users of the previous official version (2.00.12). If you would like to contribute by addressing any of the outstanding [Issues](https://github.com/OpenWaterAnalytics/EPANET/issues), then please comment on the Issue, then Fork this repo to your own account and base your commits on the [`dev` branch](https://github.com/OpenWaterAnalytics/EPANET/tree/dev). Once you are finished, you can open a Pull Request to test the code and discuss merging your changes back into the community respository.
15
15
16
-
A step-by-step tutorial on how to contribute to EPANET using GitHub is also [available](http://www.slideshare.net/demetriseliades/contributing-to-epanet-using-github-in-windows).
16
+
A step-by-step tutorial on how to contribute to EPANET using GitHub is also [available](http://www.slideshare.net/demetriseliades/contributing-to-epanet-using-github-in-windows).
17
17
18
18
__Note:__ This repository is not affiliated with, or endorsed by, the USEPA. For the last "official" release of EPANET (2.00.12 UI and Toolkit) please go to the [EPA's GitHub repo](https://github.com/USEPA/Water-Distribution-Network-Model) or [the USEPA website](http://www2.epa.gov/water-research/epanet). It is also not the graphical user interface version. This is the hydraulic and water quality solver engine.
19
19
20
20
However, if you are interested in extending EPANET for academic, personal, or commercial use, then you've come to the right place. For community discussion, FAQ, and roadmapping of the project, go to the [Community Forum](http://community.wateranalytics.org/category/epanet).
This document describes the changes and updates that have been made to version 2.2 of EPANET.
8
+
9
+
## Thread-Safe API Functions
10
+
11
+
A duplicate set of the version 2.1 API functions has been provided that allow multiple EPANET projects to be analyzed concurrently in a thread-safe manner. These functions maintain the same name as the original but use a `EN_` prefix instead of `EN`. In addition, the first argument to each of these functions is a pointer to an `EN_Project` structure that encapsulates the network data for the particular project being analyzed. For example, instead of writing:
Two new functions have been added to the API to manage the creation and deletion of project pointers. `EN_createproject` creates a new project along with a pointer to it, while `EN_deleteproject` deletes a project. An example of using the thread-safe version of the API is shown below:
22
+
```
23
+
#include "epanet2.h"
24
+
int runEpanet(char *finp, char *frpt)
25
+
{
26
+
EN_Project *pr = NULL;
27
+
int err;
28
+
err = EN_createproject(&pr);
29
+
if (err) return err;
30
+
err = EN_open(pr, finp, frpt, "");
31
+
if (!err) err = EN_solveH(pr);
32
+
if (!err) err = EN_report(pr);
33
+
EN_close(pr);
34
+
EN_deleteproject(pr);
35
+
return err;
36
+
}
37
+
```
38
+
39
+
## Additional Convergence Parameters
40
+
41
+
Two new analysis options have been added to provide more rigorous convergence criteria for EPANET's hydraulic solver. In the API they are named `EN_HEADERROR` and `EN_FLOWCHANGE` while in the `[OPTIONS]` section of an EPANET input file they are named `HEADERROR` and `FLOWCHANGE`, respectively.
42
+
43
+
`EN_HEADERROR` is the maximum head loss error that any network link can have for hydraulic convergence to occur. A link's head loss error is the difference between the head loss found as a function of computed flow in the link (such as by the Hazen-Williams equation for a pipe) and the difference in computed heads for the link's end nodes. The units of this parameter are feet (or meters for SI units). The default value of 0 indicates that no head error limit applies.
44
+
45
+
`EN_FLOWCHANGE` is the largest change in flow that any network element (link, emitter, or pressure-dependent demand) can have for hydraulic convergence to occur. It is specified in whatever flow units the project is using. The default value of 0 indicates that no flow change limit applies.
46
+
47
+
These new parameters augment the current `EN_ACCURACY` option which always remains in effect. In addition, both `EN_HEADERROR` and `EN_FLOWCHANGE` can be used as parameters in the `ENgetstatistic` (or `EN_getstatistic`) function to retrieve their computed values (even when their option values are 0) after a hydraulic solution has been completed.
48
+
49
+
## Improved Linear Solver Routine
50
+
EPANET's hydraulic solver requires solving a system of linear equations over a series of iterations until a set of convergence criteria are met. The coefficient matrix of this linear system is square and symmetric. It has a row for each network node and a non-zero off-diagonal coefficient for each link. The numerical effort needed to solve the linear system can be reduced if the nodes are re-ordered so that the non-zero coefficients cluster more tightly around the diagonal.
51
+
52
+
EPANET's original node re-ordering scheme has been replaced by the more powerful **Multiple Minimum Degree (MMD)** algorithm. On a series of eight networks ranging in size from 7,700 to 100,000 nodes **MMD** reduced the solution time for a single period (steady state) hydraulic analysis by an average of more than 50%.
53
+
54
+
## Pressure Dependent Demands
55
+
EPANET has always employed a Demand Driven Analysis (**DDA**) when modeling network hydraulics. Under this approach nodal demands at a given point in time are fixed values that must be delivered no matter what nodal heads and link flows are produced by a hydraulic solution. This can result in situations where required demands are satisfied at nodes that have negative pressures - a physical impossibility.
56
+
57
+
To address this issue EPANET has been extended to use a Pressure Driven Analysis (**PDA**) if so desired. Under **PDA**, the demand *D* delivered at a node depends on the node's available pressure *P* according to:
58
+
$$D =D_f\left(\frac{P-P_{min}}{P_{req}-P_{min}}\right)^{P_{exp}} for P_{0}<=P<=P_{req}$$where *D<sub>f</sub>* is the full demand required, *P<sub>min</sub>* is the pressure below which demand is zero, *P<sub>req</sub>* is the pressure required to deliver the full required demand and *P<sub>exp</sub>* is an exponent. When *P < P<sub>min</sub>* demand is 0 and when *P > P<sub>req</sub>* demand equals *D<sub>f</sub>*.
59
+
60
+
To implement pressure dependent analysis four new parameters have been added to the [OPTIONS] section of the EPANET input file:
61
+
62
+
63
+
| Parameter | Description | Default |
64
+
|--|--|--|
65
+
| DEMAND MODEL | either DDA or PDA | DDA |
66
+
| MINIMUM PRESSURE | value for *P<sub>min</sub>* | 0
67
+
| REQUIRED PRESSURE | value for *P<sub>req</sub>* | 0
68
+
| PRESSURE EXPONENT | value for *P<sub>exp</sub>*| 0.5 |
69
+
70
+
These parameters can also be set and retrieved in code using the following API functions
71
+
```
72
+
int ENsetdemandmodel(int modelType, double pMin, double pReq, double pExp);
73
+
int ENgetdemandmodel(int *modelType, double *pMin, double *pReq, double *pExp);
74
+
```
75
+
for the legacy API and
76
+
```
77
+
int EN_setdemandmodel(EN_Project *pr, int modelType, double pMin, double pReq, double pExp);
78
+
int EN_getdemandmodel(EN_Project *pr, int *modelType, double *pMin, double *pReq, double *pExp);
79
+
```
80
+
for the thread-safe API. Some additional points regarding the new **PDA** option are:
81
+
82
+
- If no DEMAND MODEL and its parameters are specified then the analysis defaults to being demand driven (**DDA**).
83
+
- This implementation of **PDA** assumes that the same parameters apply to all nodes in the network. Extending the framework to allow different parameters for specific nodes is straightforward to do but is left as a future feature to implement.
84
+
-*P<sub>0</sub>* is allowed to equal to *P<sub>req</sub>*. This condition can be used to find a solution that results in the smallest amount of demand reductions needed to insure that no node delivers positive demand at a pressure below *P<sub>min</min>*.
85
+
86
+
## Code Changes
87
+
88
+
- The header file `vars.h` containing global variables has been eliminated. Instead a number of new structures incorporating these variables has been added to `types.h`. These structures have been incorporated into the new `EN_Project` structure, also defined in `types.h`, which gets passed into each of the thread-safe API functions as a pointer.
89
+
- Each of the legacy API functions now simply calls its thread-safe counterpart passing in a pointer to a default global`EN_Project` variable that is declared in `types.h`.
90
+
- Throughout all code modules, global variables that were previously accessed through `vars.h` are now accessed using the `EN_Project` pointer that is passed into the functions where the variables appear.
91
+
- The exceedingly long `hydraul.c` file has been split into four separate files:
92
+
-`hydraul.c` now contains just the code needed to initialize a hydraulic analysis, set demands and control actions at each time step, and determine the length of the next time step to take.
93
+
-`hydsolver.c` implements EPANET's hydraulic solver at a single point in time.
94
+
-`hydcoeffs.c` computes values of the matrix coefficients (derived from link head losses and their gradients) used by the hydraulic solver.
95
+
-`hydstatus.c` checks for status changes in valves and pumps as requested by the hydraulic solver.
96
+
- The Multiple Minimum Degree re-ordering algorithm appears in a new file named `genmmd.c`. This is 1990's legacy code that is readily available on the web and can be found in several linear equation solver libraries.
0 commit comments